id
int64 599M
3.29B
| url
stringlengths 58
61
| html_url
stringlengths 46
51
| number
int64 1
7.72k
| title
stringlengths 1
290
| state
stringclasses 2
values | comments
int64 0
70
| created_at
timestamp[s]date 2020-04-14 10:18:02
2025-08-05 09:28:51
| updated_at
timestamp[s]date 2020-04-27 16:04:17
2025-08-05 11:39:56
| closed_at
timestamp[s]date 2020-04-14 12:01:40
2025-08-01 05:15:45
⌀ | user_login
stringlengths 3
26
| labels
listlengths 0
4
| body
stringlengths 0
228k
⌀ | is_pull_request
bool 2
classes |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2,367,890,622
|
https://api.github.com/repos/huggingface/datasets/issues/6992
|
https://github.com/huggingface/datasets/issues/6992
| 6,992
|
Dataset with streaming doesn't work with proxy
|
open
| 1
| 2024-06-22T16:12:08
| 2024-06-25T15:43:05
| null |
YHL04
|
[] |
### Describe the bug
I'm currently trying to stream data using dataset since the dataset is too big but it hangs indefinitely without loading the first batch. I use AIMOS which is a supercomputer that uses proxy to connect to the internet. I assume it has to do with the network configurations. I've already set up both HTTP_PROXY and HTTPS_PROXY. streaming = False works fine.
### Steps to reproduce the bug
use load_dataset with streaming = True in AIMOS
### Expected behavior
does not hang indefinitely and loads batches to start training run
### Environment info
_libgcc_mutex 0.1 conda_forge conda-forge
_openmp_mutex 4.5 2_gnu conda-forge
_pytorch_select 2.0 cuda_2 https://ftp.osuosl.org/pub/open-ce/1.10.0
abseil-cpp 20220623.0 h9888cd1_6 conda-forge
absl-py 1.0.0 py311h399429b_0 https://ftp.osuosl.org/pub/open-ce/1.10.0
aiofiles 23.2.1 pyhd8ed1ab_0 conda-forge
aiohttp 3.8.6 py311hf118e41_0
aiosignal 1.2.0 pyhd3eb1b0_0
archspec 0.2.3 pyhd8ed1ab_0 conda-forge
arrow-cpp 11.0.0 ha3edaa6_5_cpu conda-forge
async-timeout 4.0.2 py311h6ffa863_0
attrs 23.1.0 py311h6ffa863_0
av 10.0.0 py311he6153ed_2 https://ftp.osuosl.org/pub/open-ce/1.10.0
aws-c-auth 0.6.24 hb81f6d7_5 conda-forge
aws-c-cal 0.5.20 h3c2b4d9_6 conda-forge
aws-c-common 0.8.11 h4194056_0 conda-forge
aws-c-compression 0.2.16 ha19333d_3 conda-forge
aws-c-event-stream 0.2.18 h12a9399_6 conda-forge
aws-c-http 0.7.4 ha2cde00_2 conda-forge
aws-c-io 0.13.17 h9189062_2 conda-forge
aws-c-mqtt 0.8.6 h40d1a04_6 conda-forge
aws-c-s3 0.2.4 hbdbe4f0_3 conda-forge
aws-c-sdkutils 0.1.7 ha19333d_3 conda-forge
aws-checksums 0.1.14 ha19333d_3 conda-forge
aws-crt-cpp 0.19.7 hd018011_7 conda-forge
aws-sdk-cpp 1.10.57 hb9575ba_4 conda-forge
blas 1.0 openblas
blinker 1.8.2 pyhd8ed1ab_0 conda-forge
boltons 23.0.0 py311h6ffa863_0
boost-cpp 1.82.0 h25e6d66_2
bottleneck 1.3.5 py311h34f6284_0
brotli 1.0.9 hf118e41_7
brotli-bin 1.0.9 hf118e41_7
brotli-python 1.0.9 py311h4a02239_7
bzip2 1.0.8 h7b6447c_0
c-ares 1.19.1 hf118e41_0
ca-certificates 2024.6.2 h0f6029e_0 conda-forge
cachetools 5.3.3 pyhd8ed1ab_0 conda-forge
certifi 2024.6.2 pyhd8ed1ab_0 conda-forge
cffi 1.15.1 py311hf118e41_3
charset-normalizer 2.0.4 pyhd3eb1b0_0
click 8.1.7 unix_pyh707e725_0 conda-forge
conda 24.5.0 py311h1af927a_0 conda-forge
conda-content-trust 0.2.0 py311h6ffa863_0
conda-libmamba-solver 23.11.1 py311h6ffa863_0
conda-package-handling 2.2.0 py311h6ffa863_0
conda-package-streaming 0.9.0 py311h6ffa863_0
contourpy 1.0.5 py311h25e6d66_0
cryptography 41.0.3 py311hb0e80e7_0
cudatoolkit 11.8.0 hedcfb66_13 conda-forge
cudnn 8.9.2_11.8 h9ceb136_1 https://ftp.osuosl.org/pub/open-ce/1.10.0
cycler 0.11.0 pyhd3eb1b0_0
datasets 2.12.0 py311h6ffa863_0
dill 0.3.6 py311h6ffa863_0
distro 1.9.0 pyhd8ed1ab_0 conda-forge
ffmpeg 4.2.2 opence_0 https://ftp.osuosl.org/pub/open-ce/1.10.0
filelock 3.9.0 py311h6ffa863_0
fmt 9.1.0 h25e6d66_0
fonttools 4.25.0 pyhd3eb1b0_0
freetype 2.12.1 hd23a775_0
frozendict 2.4.4 py311hb02d432_0 conda-forge
frozenlist 1.4.0 py311hf118e41_0
fsspec 2023.9.2 py311h6ffa863_0
gflags 2.2.2 he6710b0_0
giflib 5.2.1 hf118e41_3
glog 0.6.0 hbe088e0_0 conda-forge
gmp 6.3.0 h46f38da_0 conda-forge
gmpy2 2.1.5 py311h2758da7_1 conda-forge
google-auth 2.30.0 pyhff2d567_0 conda-forge
google-auth-oauthlib 0.5.3 pyhd8ed1ab_0 conda-forge
grpc-cpp 1.51.1 h8ba971d_1 conda-forge
grpcio 1.54.3 py311h414e0d3_0 https://ftp.osuosl.org/pub/open-ce/1.10.0
huggingface_hub 0.17.3 py311h6ffa863_0
icu 73.1 h4a02239_0
idna 3.4 py311h6ffa863_0
importlib-metadata 6.0.0 py311h6ffa863_0
jinja2 3.1.4 pyhd8ed1ab_0 conda-forge
jpeg 9e hf118e41_1
jsonpatch 1.32 pyhd3eb1b0_0
jsonpointer 2.1 pyhd3eb1b0_0
kiwisolver 1.4.4 py311h4a02239_0
krb5 1.20.1 hc019ccd_1
lame 3.100 hb283c62_1003 conda-forge
lcms2 2.12 h2045e0b_0
ld_impl_linux-ppc64le 2.38 hec883e6_1
lerc 3.0 h29c3540_0
leveldb 1.23 h24532b4_1 conda-forge
libabseil 20220623.0 cxx17_h9235812_6 conda-forge
libarchive 3.6.2 hd8ab008_2
libarrow 11.0.0 h837770b_5_cpu conda-forge
libboost 1.82.0 haf51a6a_2
libbrotlicommon 1.0.9 hf118e41_7
libbrotlidec 1.0.9 hf118e41_7
libbrotlienc 1.0.9 hf118e41_7
libcrc32c 1.1.2 h3b9df90_0 conda-forge
libcurl 8.4.0 h4d62439_0
libdeflate 1.17 hf118e41_1
libedit 3.1.20221030 hf118e41_0
libev 4.33 h140841e_1
libevent 2.1.10 h19c23f1_4 conda-forge
libexpat 2.6.2 h46f38da_0 conda-forge
libffi 3.4.4 h4a02239_0
libgcc-ng 13.2.0 h31e42bb_10 conda-forge
libgfortran-ng 11.2.0 hb3889a9_1
libgfortran5 11.2.0 h1234567_1
libgomp 13.2.0 h31e42bb_10 conda-forge
libgoogle-cloud 2.7.0 h11140b6_1 conda-forge
libgrpc 1.51.1 h4d29a31_1 conda-forge
libmamba 1.5.3 h7c6fafd_0
libmambapy 1.5.3 py311h828bf7b_0
libnghttp2 1.57.0 h44e5816_0
libnsl 2.0.1 ha17a0cc_0 conda-forge
libopenblas 0.3.23 hc5a31fb_2 https://ftp.osuosl.org/pub/open-ce/1.10.0
libopus 1.3.1 h4e0d66e_1 conda-forge
libpng 1.6.39 hf118e41_0
libprotobuf 3.21.12 h1776448_0 https://ftp.osuosl.org/pub/open-ce/1.10.0
libsolv 0.7.24 h0f529ac_0
libsqlite 3.45.3 hd4bbf49_0 conda-forge
libssh2 1.10.0 h50fa78f_2
libstdcxx-ng 13.2.0 h262982c_10 conda-forge
libthrift 0.18.0 h82f1162_0 conda-forge
libtiff 4.5.1 h4a02239_0
libutf8proc 2.8.0 hb283c62_0 conda-forge
libuuid 2.38.1 h4194056_0 conda-forge
libvpx 1.13.1 h46f38da_0 conda-forge
libwebp 1.3.2 h0f96ee2_0
libwebp-base 1.3.2 hf118e41_0
libxcrypt 4.4.36 ha17a0cc_1 conda-forge
libxml2 2.10.4 h18e3229_1
libzlib 1.2.13 h1f2b957_6 conda-forge
llvm-openmp 14.0.6 hc028133_0 https://ftp.osuosl.org/pub/open-ce/1.10.0
lmdb 0.9.31 ha17a0cc_1 conda-forge
lz4-c 1.9.4 h4a02239_0
markdown 3.4.4 pyhd8ed1ab_0 conda-forge
markupsafe 2.1.5 py311h32d8acf_0 conda-forge
matplotlib 3.8.0 py311h6ffa863_0
matplotlib-base 3.8.0 py311h52e1fcc_0
menuinst 2.1.1 py311h1af927a_0 conda-forge
mpc 1.3.1 heaf1863_0 conda-forge
mpfr 4.2.1 haad2271_1 conda-forge
mpmath 1.3.0 pyhd8ed1ab_0 conda-forge
multidict 6.0.2 py311hf118e41_0
multiprocess 0.70.14 py311h6ffa863_0
munkres 1.1.4 py_0
mypy_extensions 1.0.0 pyha770c72_0 conda-forge
nccl 2.18.3 cuda11.8_1 https://ftp.osuosl.org/pub/open-ce/1.10.0
ncurses 6.4 h4a02239_0
nest-asyncio 1.6.0 pyhd8ed1ab_0 conda-forge
networkx 2.8.8 pyhd8ed1ab_0 conda-forge
nomkl 3.0 0 https://ftp.osuosl.org/pub/open-ce/1.10.0
numactl 2.0.16 hba61f60_1 https://ftp.osuosl.org/pub/open-ce/1.10.0
numexpr 2.8.7 py311hc46fc55_0
numpy 1.24.3 py311h148a09e_0
numpy-base 1.24.3 py311h06b82f6_0
oauthlib 3.2.2 pyhd8ed1ab_0 conda-forge
openjpeg 2.4.0 hfe35807_0
openssl 3.3.1 h1f2b957_0 conda-forge
orc 1.8.2 h341c9a4_2 conda-forge
packaging 23.1 py311h6ffa863_0
pandas 2.1.1 py311h52e1fcc_0
pcre2 10.42 h280155c_0
pillow 10.0.1 py311he33076b_0
pip 23.3 py311h6ffa863_0
platformdirs 4.2.2 pyhd8ed1ab_0 conda-forge
pluggy 1.0.0 py311h6ffa863_1
pooch 1.8.2 pyhd8ed1ab_0 conda-forge
protobuf 4.21.12 py311ha7baec7_1 https://ftp.osuosl.org/pub/open-ce/1.10.0
psutil 5.9.8 py311hd26027c_0 conda-forge
pyarrow 11.0.0 py311h04a18d5_1
pyasn1 0.6.0 pyhd8ed1ab_0 conda-forge
pyasn1-modules 0.4.0 pyhd8ed1ab_0 conda-forge
pybind11-abi 4 hd3eb1b0_1
pycosat 0.6.6 py311hf118e41_0
pycparser 2.21 pyhd3eb1b0_0
pyjwt 2.8.0 pyhd8ed1ab_1 conda-forge
pyopenssl 23.2.0 py311h6ffa863_0
pyparsing 3.0.9 py311h6ffa863_0
pyre-extensions 0.0.30 pyhd8ed1ab_0 conda-forge
pysocks 1.7.1 py311h6ffa863_0
python 3.11.8 h3332dee_0_cpython conda-forge
python-dateutil 2.8.2 pyhd3eb1b0_0
python-tzdata 2023.3 pyhd3eb1b0_0
python-xxhash 2.0.2 py311hf118e41_1
python_abi 3.11 4_cp311 conda-forge
pytorch 2.0.1 cuda11.8_py311_1 https://ftp.osuosl.org/pub/open-ce/1.10.0
pytorch-base 2.0.1 cuda11.8_py311_pb4.21.12_4 https://ftp.osuosl.org/pub/open-ce/1.10.0
pytz 2023.3.post1 py311h6ffa863_0
pyu2f 0.1.5 pyhd8ed1ab_0 conda-forge
pyyaml 6.0.1 py311hf118e41_0
re2 2023.02.01 h883269e_0 conda-forge
readline 8.2 hf118e41_0
regex 2023.10.3 py311hf118e41_0
reproc 14.2.4 h29c3540_1
reproc-cpp 14.2.4 h29c3540_1
requests 2.31.0 py311h6ffa863_0
requests-oauthlib 2.0.0 pyhd8ed1ab_0 conda-forge
responses 0.13.3 pyhd3eb1b0_0
rsa 4.9 pyhd8ed1ab_0 conda-forge
ruamel.yaml 0.17.21 py311hf118e41_0
s2n 1.3.37 h5e47323_0 conda-forge
safetensors 0.4.0 py311hda16d9e_0
scipy 1.11.1 py311hd69e9bb_0 https://ftp.osuosl.org/pub/open-ce/1.10.0
sentencepiece 0.1.97 h1e74c73_py311_pb4.21.12_2 https://ftp.osuosl.org/pub/open-ce/1.10.0
setuptools 68.0.0 py311h6ffa863_0
six 1.16.0 pyhd3eb1b0_1
snappy 1.1.9 h29c3540_0
sqlite 3.41.2 hf118e41_0
sympy 1.12.1 pypyh2585a3b_103 conda-forge
tabulate 0.8.10 pyhd8ed1ab_0 conda-forge
tensorboard 2.13.0 pyhab0730d_pb4.21.12_1 https://ftp.osuosl.org/pub/open-ce/1.10.0
tensorboard-data-server 0.7.0 pyh6f84499_1 https://ftp.osuosl.org/pub/open-ce/1.10.0
tensorboard-plugin-wit 1.6.0 pyh9f0ad1d_0 conda-forge
tk 8.6.13 hd4bbf49_0 conda-forge
tokenizers 0.13.3 py311h3d4f45a_0
torchdata 0.6.0 py311_2 https://ftp.osuosl.org/pub/open-ce/1.10.0
torchsnapshot 0.1.0 pyhd8ed1ab_0 conda-forge
torchtext-base 0.15.2 cuda11.8_py311_1 https://ftp.osuosl.org/pub/open-ce/1.10.0
torchtnt 0.2.4 pyhd8ed1ab_0 conda-forge
torchvision-base 0.15.2 cuda11.8_py311_1 https://ftp.osuosl.org/pub/open-ce/1.10.0
tornado 6.3.3 py311hf118e41_0
tqdm 4.65.0 py311h7837921_0
transformers 4.32.1 py311h6ffa863_0
truststore 0.8.0 py311h6ffa863_0
typing-extensions 4.7.1 py311h6ffa863_0
typing_extensions 4.7.1 py311h6ffa863_0
typing_inspect 0.9.0 pyhd8ed1ab_0 conda-forge
tzdata 2023c h04d1e81_0
urllib3 1.26.18 py311h6ffa863_0
utf8proc 2.6.1 h140841e_0
werkzeug 2.3.8 pyhd8ed1ab_0 conda-forge
wheel 0.41.2 py311h6ffa863_0
xxhash 0.8.0 h140841e_3
xz 5.4.2 hf118e41_0
yaml 0.2.5 h7b6447c_0
yaml-cpp 0.8.0 h4a02239_0
yarl 1.8.1 py311hf118e41_0
zipp 3.11.0 py311h6ffa863_0
zlib 1.2.13 h1f2b957_6 conda-forge
zstandard 0.19.0 py311hf118e41_0
zstd 1.5.5 h57e4825_0
| false
|
2,367,711,094
|
https://api.github.com/repos/huggingface/datasets/issues/6991
|
https://github.com/huggingface/datasets/pull/6991
| 6,991
|
Unblock NumPy 2.0
|
closed
| 21
| 2024-06-22T09:19:53
| 2024-12-25T17:57:34
| 2024-07-12T12:04:53
|
NeilGirdhar
|
[] |
Fixes https://github.com/huggingface/datasets/issues/6980
| true
|
2,366,660,785
|
https://api.github.com/repos/huggingface/datasets/issues/6990
|
https://github.com/huggingface/datasets/issues/6990
| 6,990
|
Problematic rank after calling `split_dataset_by_node` twice
|
closed
| 1
| 2024-06-21T14:25:26
| 2024-06-25T16:19:19
| 2024-06-25T16:19:19
|
yzhangcs
|
[] |
### Describe the bug
I'm trying to split `IterableDataset` by `split_dataset_by_node`.
But when doing split on a already split dataset, the resulting `rank` is greater than `world_size`.
### Steps to reproduce the bug
Here is the minimal code for reproduction:
```py
>>> from datasets import load_dataset
>>> from datasets.distributed import split_dataset_by_node
>>> dataset = load_dataset('fla-hub/slimpajama-test', split='train', streaming=True)
>>> dataset = split_dataset_by_node(dataset, 1, 32)
>>> dataset._distributed
DistributedConfig(rank=1, world_size=32)
>>> dataset = split_dataset_by_node(dataset, 1, 15)
>>> dataset._distributed
DistributedConfig(rank=481, world_size=480)
```
As you can see, the second rank 481 > 480, which is problematic.
### Expected behavior
I think this error comes from this line @lhoestq
https://github.com/huggingface/datasets/blob/a6ccf944e42c1a84de81bf326accab9999b86c90/src/datasets/iterable_dataset.py#L2943-L2944
We may need to obtain the rank first. Then the above code gives
```py
>>> dataset._distributed
DistributedConfig(rank=16, world_size=480)
```
### Environment info
datasets==2.20.0
| false
|
2,365,556,449
|
https://api.github.com/repos/huggingface/datasets/issues/6989
|
https://github.com/huggingface/datasets/issues/6989
| 6,989
|
cache in nfs error
|
open
| 1
| 2024-06-21T02:09:22
| 2025-01-29T11:44:04
| null |
simplew2011
|
[] |
### Describe the bug
- When reading dataset, a cache will be generated to the ~/. cache/huggingface/datasets directory
- When using .map and .filter operations, runtime cache will be generated to the /tmp/hf_datasets-* directory
- The default is to use the path of tempfile.tempdir
- If I modify this path to the NFS disk, an error will be reported, but the program will continue to run
- https://github.com/huggingface/datasets/blob/main/src/datasets/config.py#L257
```
Traceback (most recent call last):
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/site-packages/multiprocess/process.py", line 315, in _bootstrap
self.run()
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/site-packages/multiprocess/process.py", line 108, in run
self._target(*self._args, **self._kwargs)
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/site-packages/multiprocess/managers.py", line 616, in _run_server
server.serve_forever()
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/site-packages/multiprocess/managers.py", line 182, in serve_forever
sys.exit(0)
SystemExit: 0
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/site-packages/multiprocess/util.py", line 300, in _run_finalizers
finalizer()
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/site-packages/multiprocess/util.py", line 224, in __call__
res = self._callback(*self._args, **self._kwargs)
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/site-packages/multiprocess/util.py", line 133, in _remove_temp_dir
rmtree(tempdir)
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/shutil.py", line 718, in rmtree
_rmtree_safe_fd(fd, path, onerror)
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/shutil.py", line 675, in _rmtree_safe_fd
onerror(os.unlink, fullname, sys.exc_info())
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/shutil.py", line 673, in _rmtree_safe_fd
os.unlink(entry.name, dir_fd=topfd)
OSError: [Errno 16] Device or resource busy: '.nfs000000038330a012000030b4'
Traceback (most recent call last):
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/site-packages/multiprocess/process.py", line 315, in _bootstrap
self.run()
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/site-packages/multiprocess/process.py", line 108, in run
self._target(*self._args, **self._kwargs)
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/site-packages/multiprocess/managers.py", line 616, in _run_server
server.serve_forever()
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/site-packages/multiprocess/managers.py", line 182, in serve_forever
sys.exit(0)
SystemExit: 0
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/site-packages/multiprocess/util.py", line 300, in _run_finalizers
finalizer()
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/site-packages/multiprocess/util.py", line 224, in __call__
res = self._callback(*self._args, **self._kwargs)
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/site-packages/multiprocess/util.py", line 133, in _remove_temp_dir
rmtree(tempdir)
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/shutil.py", line 718, in rmtree
_rmtree_safe_fd(fd, path, onerror)
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/shutil.py", line 675, in _rmtree_safe_fd
onerror(os.unlink, fullname, sys.exc_info())
File "/home/wzp/miniconda3/envs/dask/lib/python3.8/shutil.py", line 673, in _rmtree_safe_fd
os.unlink(entry.name, dir_fd=topfd)
OSError: [Errno 16] Device or resource busy: '.nfs0000000400064d4a000030e5'
```
### Steps to reproduce the bug
```
import os
import time
import tempfile
from datasets import load_dataset
def add_column(sample):
# print(type(sample))
# time.sleep(0.1)
sample['__ds__stats__'] = {'data': 123}
return sample
def filt_column(sample):
# print(type(sample))
if len(sample['content']) > 10:
return True
else:
return False
if __name__ == '__main__':
input_dir = '/mnt/temp/CN/small' # some json dataset
dataset = load_dataset('json', data_dir=input_dir)
temp_dir = '/media/release/release/temp/temp' # a nfs folder
os.makedirs(temp_dir, exist_ok=True)
# change huggingface-datasets runtime cache in nfs（default in /tmp）
tempfile.tempdir = temp_dir
aa = dataset.map(add_column, num_proc=64)
aa = aa.filter(filt_column, num_proc=64)
print(aa)
```
### Expected behavior
no error occur
### Environment info
datasets==2.18.0
ubuntu 20.04
| false
|
2,364,129,918
|
https://api.github.com/repos/huggingface/datasets/issues/6988
|
https://github.com/huggingface/datasets/pull/6988
| 6,988
|
[`feat`] Move dataset card creation to method for easier overriding
|
open
| 6
| 2024-06-20T10:47:57
| 2024-06-21T16:04:58
| null |
tomaarsen
|
[] |
Hello!
## Pull Request overview
* Move dataset card creation to method for easier overriding
## Details
It's common for me to fully automatically download, reformat, and upload a dataset (e.g. see https://huggingface.co/datasets?other=sentence-transformers), but one aspect that I cannot easily automate is the dataset card generation. This is because during `push_to_hub`, the dataset card is created in 3 lines of code in a much larger method. To automatically generate a dataset card, I need to either:
1. Subclass `Dataset`/`DatasetDict`, copy the entire `push_to_hub` method to override the ~3 lines used to generate the dataset card. This is not viable as the method is likely to change over time.
2. Use `push_to_hub` normally, then separately download the pushed (but empty) dataset card, update it, and reupload the modified dataset. This works fine, but prevents me from being able to return a `Dataset` to my users which will automatically use a nice dataset card.
So, in this PR I'm proposing to move the dataset generation into another method so that it can be overridden more easily. For example, imagine the following use case:
````python
import json
from typing import Any, Dict, Optional
from datasets import Dataset, load_dataset
from datasets.info import DatasetInfosDict, DatasetInfo
from datasets.utils.metadata import MetadataConfigs
from huggingface_hub import DatasetCardData, DatasetCard
TEMPLATE = r"""---
{dataset_card_data}
---
# Dataset Card for {source_dataset_name} with mined hard negatives
This dataset is a collection of {column_one}-{column_two}-negative triplets from the {source_dataset_name} dataset. See [{source_dataset_name}](https://huggingface.co/datasets/{source_dataset_id}) for additional information. This dataset can be used directly with Sentence Transformers to train embedding models.
## Mining Parameters
The negative samples have been mined using the following parameters:
- `range_min`: {range_min}, i.e. we skip the {range_min} most similar samples
- `range_max`: {range_max}, i.e. we only look at the top {range_max} most similar samples
- `margin`: {margin}, i.e. we require negative similarity + margin < positive similarity, so negative samples can't be more similar than the known true answer
- `sampling_strategy`: {sampling_strategy}, i.e. whether to randomly sample from the candidate negatives or take the "top" negatives
- `num_negatives`: {num_negatives}, i.e. we mine {num_negatives} negatives per question-answer pair
## Dataset Format
- Columns: {column_one}, {column_two}, negative
- Column types: str, str, str
- Example:
```python
{example}
```
"""
class HNMDataset(Dataset):
@classmethod
def from_dict(cls, *args, mining_kwargs: Dict[str, Any], **kwargs) -> "HNMDataset":
dataset = super().from_dict(*args, **kwargs)
dataset.mining_kwargs = mining_kwargs
return dataset
def _create_dataset_card(
self,
dataset_card_data: DatasetCardData,
dataset_card: Optional[DatasetCard],
config_name: str,
info_to_dump: DatasetInfo,
metadata_config_to_dump: MetadataConfigs,
) -> DatasetCard:
if dataset_card:
return dataset_card
DatasetInfosDict({config_name: info_to_dump}).to_dataset_card_data(dataset_card_data)
MetadataConfigs({config_name: metadata_config_to_dump}).to_dataset_card_data(dataset_card_data)
dataset_card_data.tags = ["sentence-transformers"]
dataset_name = self.mining_kwargs["source_dataset"].info.dataset_name
# Very messy, just as an example:
dataset_id = list(self.mining_kwargs["source_dataset"].info.download_checksums.keys())[0].removeprefix("hf://datasets/").split("@")[0]
content = TEMPLATE.format(**{
"dataset_card_data": str(dataset_card_data),
"source_dataset_name": dataset_name,
"source_dataset_id": dataset_id,
"range_min": self.mining_kwargs["range_min"],
"range_max": self.mining_kwargs["range_max"],
"margin": self.mining_kwargs["margin"],
"sampling_strategy": self.mining_kwargs["sampling_strategy"],
"num_negatives": self.mining_kwargs["num_negatives"],
"column_one": self.column_names[0],
"column_two": self.column_names[1],
"example": json.dumps(self[0], indent=4),
})
return DatasetCard(content)
source_dataset = load_dataset("sentence-transformers/gooaq", split="train[:100]")
dataset = HNMDataset.from_dict({
"query": source_dataset["question"],
"answer": source_dataset["answer"],
# "negative": ... <- In my case, this column would be 'mined' automatically with these parameters
}, mining_kwargs={
"range_min": 10,
"range_max": 20,
"max_score": 0.9,
"margin": 0.1,
"sampling_strategy": "random",
"num_negatives": 3,
"source_dataset": source_dataset,
})
dataset.push_to_hub("tomaarsen/mining_demo", private=True)
````
In this script, I've created a subclass which stores some additional information about how the dataset was generated. It's a bit hacky (e.g. setting a `mining_kwargs` parameter in `from_dict` that wasn't created in `__init__`, but that's just a consequence of how the `from_...` methods don't accept kwargs), but it allows me to create a "hard negatives mining" function that returns a dataset which people can use locally like normal, but if they choose to upload it, then it'll automatically include some information, e.g.: https://huggingface.co/datasets/tomaarsen/mining_demo
This allows others to actually find this dataset (e.g. via the `sentence-transformers` tag) and get an idea of the quality, source, etc. by looking at the model card.
## Note
I'm not fixed on this solution whatsoever: I am also completely fine with other solutions, e.g. a `dataset.set_dataset_card_creator` method that allows me to provide a function without even having to subclass anything. I'm open to all ideas :)
cc @albertvillanova @lhoestq
cc @LysandreJik
- Tom Aarsen
| true
|
2,363,728,190
|
https://api.github.com/repos/huggingface/datasets/issues/6987
|
https://github.com/huggingface/datasets/pull/6987
| 6,987
|
Remove beam
|
closed
| 2
| 2024-06-20T07:27:14
| 2024-06-26T19:41:55
| 2024-06-26T19:35:42
|
albertvillanova
|
[] |
Remove beam, as part of the 3.0 release.
| true
|
2,362,584,179
|
https://api.github.com/repos/huggingface/datasets/issues/6986
|
https://github.com/huggingface/datasets/pull/6986
| 6,986
|
Add large_list type support in string_to_arrow
|
closed
| 1
| 2024-06-19T14:54:25
| 2024-08-12T14:43:48
| 2024-08-12T14:43:47
|
arthasking123
|
[] |
add large_list type support in string_to_arrow() and _arrow_to_datasets_dtype() in features.py
Fix #6984
| true
|
2,362,378,276
|
https://api.github.com/repos/huggingface/datasets/issues/6985
|
https://github.com/huggingface/datasets/issues/6985
| 6,985
|
AttributeError: module 'pyarrow.lib' has no attribute 'ListViewType'
|
closed
| 14
| 2024-06-19T13:22:28
| 2025-03-14T18:47:53
| 2024-06-25T05:40:51
|
firmai
|
[] |
### Describe the bug
I have been struggling with this for two days, any help would be appreciated. Python 3.10
```
from setfit import SetFitModel
from huggingface_hub import login
access_token_read = "cccxxxccc"
# Authenticate with the Hugging Face Hub
login(token=access_token_read)
# Load the models from the Hugging Face Hub
trainer_relv = SetFitModel.from_pretrained("snowdere/trainer_relevance")
trainer_trust = SetFitModel.from_pretrained("snowdere/trainer_trust")
trainer_sent = SetFitModel.from_pretrained("snowdere/trainer_sent")
trainer_topic = SetFitModel.from_pretrained("snowdere/trainer_topic")
```
```
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
Cell In[6], line 1
----> 1 from setfit import SetFitModel
2 from huggingface_hub import login
4 access_token_read = "ccsddsds"
File /opt/conda/lib/python3.10/site-packages/setfit/__init__.py:7
4 import os
5 import warnings
----> 7 from .data import get_templated_dataset, sample_dataset
8 from .model_card import SetFitModelCardData
9 from .modeling import SetFitHead, SetFitModel
File /opt/conda/lib/python3.10/site-packages/setfit/data.py:5
3 import pandas as pd
4 import torch
----> 5 from datasets import Dataset, DatasetDict, load_dataset
6 from torch.utils.data import Dataset as TorchDataset
8 from . import logging
File /opt/conda/lib/python3.10/site-packages/datasets/__init__.py:18
1 # ruff: noqa
2 # Copyright 2020 The HuggingFace Datasets Authors and the TensorFlow Datasets Authors.
3 #
(...)
13 # See the License for the specific language governing permissions and
14 # limitations under the License.
16 __version__ = "2.19.0"
---> 18 from .arrow_dataset import Dataset
19 from .arrow_reader import ReadInstruction
20 from .builder import ArrowBasedBuilder, BeamBasedBuilder, BuilderConfig, DatasetBuilder, GeneratorBasedBuilder
File /opt/conda/lib/python3.10/site-packages/datasets/arrow_dataset.py:76
73 from tqdm.contrib.concurrent import thread_map
75 from . import config
---> 76 from .arrow_reader import ArrowReader
77 from .arrow_writer import ArrowWriter, OptimizedTypedSequence
78 from .data_files import sanitize_patterns
File /opt/conda/lib/python3.10/site-packages/datasets/arrow_reader.py:29
26 from typing import TYPE_CHECKING, List, Optional, Union
28 import pyarrow as pa
---> 29 import pyarrow.parquet as pq
30 from tqdm.contrib.concurrent import thread_map
32 from .download.download_config import DownloadConfig
File /opt/conda/lib/python3.10/site-packages/pyarrow/parquet/__init__.py:20
1 # Licensed to the Apache Software Foundation (ASF) under one
2 # or more contributor license agreements. See the NOTICE file
3 # distributed with this work for additional information
(...)
17
18 # flake8: noqa
---> 20 from .core import *
File /opt/conda/lib/python3.10/site-packages/pyarrow/parquet/core.py:33
30 import pyarrow as pa
32 try:
---> 33 import pyarrow._parquet as _parquet
34 except ImportError as exc:
35 raise ImportError(
36 "The pyarrow installation is not built with support "
37 f"for the Parquet file format ({str(exc)})"
38 ) from None
File /opt/conda/lib/python3.10/site-packages/pyarrow/_parquet.pyx:1, in init pyarrow._parquet()
AttributeError: module 'pyarrow.lib' has no attribute 'ListViewType'
```
setfit: 1.0.3
transformers: 4.41.2
lingua-language-detector: 2.0.2
polars: 0.20.31
lightning: None
google-cloud-bigquery: 3.24.0
shapely: 2.0.4
pyarrow: 16.0.0
### Steps to reproduce the bug
I have tried all version combinations for Dataset and Pyarrow, the all have the same error since a few days ago. This is accross multiple scripts I have.
### Expected behavior
Just ron normally.
### Environment info
3.10
| false
|
2,362,143,554
|
https://api.github.com/repos/huggingface/datasets/issues/6984
|
https://github.com/huggingface/datasets/issues/6984
| 6,984
|
Convert polars DataFrame back to datasets
|
closed
| 1
| 2024-06-19T11:38:48
| 2024-08-12T14:43:46
| 2024-08-12T14:43:46
|
ljw20180420
|
[
"enhancement"
] |
### Feature request
This returns error.
```python
from datasets import Dataset
dsdf = Dataset.from_dict({"x": [[1, 2], [3, 4, 5]], "y": ["a", "b"]})
Dataset.from_polars(dsdf.to_polars())
```
ValueError: Arrow type large_list<item: int64> does not have a datasets dtype equivalent.
### Motivation
When datasets contain Sequence data type, it will be converted to Arrow type large_list. However, the reverse (from large_list to Sequence) does not work.
### Your contribution
No
| false
|
2,361,806,201
|
https://api.github.com/repos/huggingface/datasets/issues/6983
|
https://github.com/huggingface/datasets/pull/6983
| 6,983
|
Remove metrics
|
closed
| 2
| 2024-06-19T09:08:55
| 2024-06-28T06:57:38
| 2024-06-28T06:51:30
|
albertvillanova
|
[] |
Remove all metrics, as part of the 3.0 release.
Note they are deprecated since 2.5.0 version.
| true
|
2,361,661,469
|
https://api.github.com/repos/huggingface/datasets/issues/6982
|
https://github.com/huggingface/datasets/issues/6982
| 6,982
|
cannot split dataset when using load_dataset
|
closed
| 3
| 2024-06-19T08:07:16
| 2024-07-08T06:20:16
| 2024-07-08T06:20:16
|
cybest0608
|
[] |
### Describe the bug
when I use load_dataset methods to load mozilla-foundation/common_voice_7_0, it can successfully download and extracted the dataset but It cannot generating the arrow document,
This bug happened in my server, my laptop, so as #6906 , but it won't happen in the google colab. I work for it for days, even I load the datasets from local path, it can Generating train split and validation split but bug happen again in test split.
### Steps to reproduce the bug
from datasets import load_dataset, load_metric, Audio
common_voice_train = load_dataset("mozilla-foundation/common_voice_7_0", "ja", split="train", token=selftoken, trust_remote_code=True)
### Expected behavior
```
{
"name": "ValueError",
"message": "Instruction \"train\" corresponds to no data!",
"stack": "---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
Cell In[2], line 3
1 from datasets import load_dataset, load_metric, Audio
----> 3 common_voice_train = load_dataset(\"mozilla-foundation/common_voice_7_0\", \"ja\", split=\"train\",token='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')#,trust_remote_code=True)#,streaming=True)
4 common_voice_test = load_dataset(\"mozilla-foundation/common_voice_7_0\", \"ja\", split=\"test\",token='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')#,trust_remote_code=True)#,streaming=True)
File c:\\Users\\cybes\\.conda\\envs\\ECoG\\lib\\site-packages\\datasets\\load.py:2626, in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, verification_mode, ignore_verifications, keep_in_memory, save_infos, revision, token, use_auth_token, task, streaming, num_proc, storage_options, trust_remote_code, **config_kwargs)
2622 # Build dataset for splits
2623 keep_in_memory = (
2624 keep_in_memory if keep_in_memory is not None else is_small_dataset(builder_instance.info.dataset_size)
2625 )
-> 2626 ds = builder_instance.as_dataset(split=split, verification_mode=verification_mode, in_memory=keep_in_memory)
2627 # Rename and cast features to match task schema
2628 if task is not None:
2629 # To avoid issuing the same warning twice
File c:\\Users\\cybes\\.conda\\envs\\ECoG\\lib\\site-packages\\datasets\\builder.py:1266, in DatasetBuilder.as_dataset(self, split, run_post_process, verification_mode, ignore_verifications, in_memory)
1263 verification_mode = VerificationMode(verification_mode or VerificationMode.BASIC_CHECKS)
1265 # Create a dataset for each of the given splits
-> 1266 datasets = map_nested(
1267 partial(
1268 self._build_single_dataset,
1269 run_post_process=run_post_process,
1270 verification_mode=verification_mode,
1271 in_memory=in_memory,
1272 ),
1273 split,
1274 map_tuple=True,
1275 disable_tqdm=True,
1276 )
1277 if isinstance(datasets, dict):
1278 datasets = DatasetDict(datasets)
File c:\\Users\\cybes\\.conda\\envs\\ECoG\\lib\\site-packages\\datasets\\utils\\py_utils.py:484, in map_nested(function, data_struct, dict_only, map_list, map_tuple, map_numpy, num_proc, parallel_min_length, batched, batch_size, types, disable_tqdm, desc)
482 if batched:
483 data_struct = [data_struct]
--> 484 mapped = function(data_struct)
485 if batched:
486 mapped = mapped[0]
File c:\\Users\\cybes\\.conda\\envs\\ECoG\\lib\\site-packages\\datasets\\builder.py:1296, in DatasetBuilder._build_single_dataset(self, split, run_post_process, verification_mode, in_memory)
1293 split = Split(split)
1295 # Build base dataset
-> 1296 ds = self._as_dataset(
1297 split=split,
1298 in_memory=in_memory,
1299 )
1300 if run_post_process:
1301 for resource_file_name in self._post_processing_resources(split).values():
File c:\\Users\\cybes\\.conda\\envs\\ECoG\\lib\\site-packages\\datasets\\builder.py:1370, in DatasetBuilder._as_dataset(self, split, in_memory)
1368 if self._check_legacy_cache():
1369 dataset_name = self.name
-> 1370 dataset_kwargs = ArrowReader(cache_dir, self.info).read(
1371 name=dataset_name,
1372 instructions=split,
1373 split_infos=self.info.splits.values(),
1374 in_memory=in_memory,
1375 )
1376 fingerprint = self._get_dataset_fingerprint(split)
1377 return Dataset(fingerprint=fingerprint, **dataset_kwargs)
File c:\\Users\\cybes\\.conda\\envs\\ECoG\\lib\\site-packages\\datasets\\arrow_reader.py:256, in BaseReader.read(self, name, instructions, split_infos, in_memory)
254 msg = f'Instruction \"{instructions}\" corresponds to no data!'
255 #msg = f'Instruction \"{self._path}\",\"{name}\",\"{instructions}\",\"{split_infos}\" corresponds to no data!'
--> 256 raise ValueError(msg)
257 return self.read_files(files=files, original_instructions=instructions, in_memory=in_memory)
ValueError: Instruction \"train\" corresponds to no data!"
}
```
### Environment info
Environment:
python 3.9
windows 11 pro
VScode+jupyter
| false
|
2,361,520,022
|
https://api.github.com/repos/huggingface/datasets/issues/6981
|
https://github.com/huggingface/datasets/pull/6981
| 6,981
|
Update docs on trust_remote_code defaults to False
|
closed
| 2
| 2024-06-19T07:12:21
| 2024-06-19T14:32:59
| 2024-06-19T14:26:37
|
albertvillanova
|
[] |
Update docs on trust_remote_code defaults to False.
The docs needed to be updated due to this PR:
- #6954
| true
|
2,360,909,930
|
https://api.github.com/repos/huggingface/datasets/issues/6980
|
https://github.com/huggingface/datasets/issues/6980
| 6,980
|
Support NumPy 2.0
|
closed
| 0
| 2024-06-18T23:30:22
| 2024-07-12T12:04:54
| 2024-07-12T12:04:53
|
NeilGirdhar
|
[
"enhancement"
] |
### Feature request
Support NumPy 2.0.
### Motivation
NumPy introduces the Array API, which bridges the gap between machine learning libraries. Many clients of HuggingFace are eager to start using the Array API.
Besides that, NumPy 2 provides a cleaner interface than NumPy 1.
### Tasks
NumPy 2.0 was released for testing so that libraries could ensure compatibility [since mid-March](https://github.com/numpy/numpy/issues/24300#issuecomment-1986815755). What needs to be done for HuggingFace to support Numpy 2?
- [x] Fix use of `array`: https://github.com/huggingface/datasets/pull/6976
- [ ] Remove [NumPy version limit](https://github.com/huggingface/datasets/pull/6975): https://github.com/huggingface/datasets/pull/6991
| false
|
2,360,175,363
|
https://api.github.com/repos/huggingface/datasets/issues/6979
|
https://github.com/huggingface/datasets/issues/6979
| 6,979
|
How can I load partial parquet files only?
|
closed
| 12
| 2024-06-18T15:44:16
| 2024-06-21T17:09:32
| 2024-06-21T13:32:50
|
lucasjinreal
|
[] |
I have a HUGE dataset about 14TB, I unable to download all parquet all. I just take about 100 from it.
dataset = load_dataset("xx/", data_files="data/train-001*-of-00314.parquet")
How can I just using 000 - 100 from a 00314 from all partially?
I search whole net didn't found a solution, **this is stupid if they didn't support it, and I swear I wont using stupid parquet any more**
| false
|
2,359,511,469
|
https://api.github.com/repos/huggingface/datasets/issues/6978
|
https://github.com/huggingface/datasets/pull/6978
| 6,978
|
Fix regression for pandas < 2.0.0 in JSON loader
|
closed
| 3
| 2024-06-18T10:26:34
| 2024-06-19T06:23:24
| 2024-06-19T05:50:18
|
albertvillanova
|
[] |
A regression was introduced for pandas < 2.0.0 in PR:
- #6914
As described in pandas docs, the `dtype_backend` parameter was first added in pandas 2.0.0: https://pandas.pydata.org/docs/reference/api/pandas.read_json.html
This PR fixes the regression by passing (or not) the `dtype_backend` parameter depending on pandas version.
Maybe, in a future 3.0 `datasets` release, we could just require pandas > 2.0.
Reported by:
- #6977
| true
|
2,359,295,045
|
https://api.github.com/repos/huggingface/datasets/issues/6977
|
https://github.com/huggingface/datasets/issues/6977
| 6,977
|
load json file error with v2.20.0
|
closed
| 2
| 2024-06-18T08:41:01
| 2024-06-18T10:06:10
| 2024-06-18T10:06:09
|
xiaoyaolangzhi
|
[] |
### Describe the bug
```
load_dataset(path="json", data_files="./test.json")
```
```
Generating train split: 0 examples [00:00, ? examples/s]
Traceback (most recent call last):
File "/usr/local/lib/python3.10/dist-packages/datasets/packaged_modules/json/json.py", line 132, in _generate_tables
pa_table = paj.read_json(
File "pyarrow/_json.pyx", line 308, in pyarrow._json.read_json
File "pyarrow/error.pxi", line 154, in pyarrow.lib.pyarrow_internal_check_status
File "pyarrow/error.pxi", line 91, in pyarrow.lib.check_status
pyarrow.lib.ArrowInvalid: JSON parse error: Column() changed from object to array in row 0
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/usr/local/lib/python3.10/dist-packages/datasets/builder.py", line 1997, in _prepare_split_single
for _, table in generator:
File "/usr/local/lib/python3.10/dist-packages/datasets/packaged_modules/json/json.py", line 155, in _generate_tables
df = pd.read_json(f, dtype_backend="pyarrow")
File "/usr/local/lib/python3.10/dist-packages/pandas/util/_decorators.py", line 211, in wrapper
return func(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/pandas/util/_decorators.py", line 331, in wrapper
return func(*args, **kwargs)
TypeError: read_json() got an unexpected keyword argument 'dtype_backend'
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/app/t1.py", line 11, in <module>
load_dataset(path=data_path, data_files="./t2.json")
File "/usr/local/lib/python3.10/dist-packages/datasets/load.py", line 2616, in load_dataset
builder_instance.download_and_prepare(
File "/usr/local/lib/python3.10/dist-packages/datasets/builder.py", line 1029, in download_and_prepare
self._download_and_prepare(
File "/usr/local/lib/python3.10/dist-packages/datasets/builder.py", line 1124, in _download_and_prepare
self._prepare_split(split_generator, **prepare_split_kwargs)
File "/usr/local/lib/python3.10/dist-packages/datasets/builder.py", line 1884, in _prepare_split
for job_id, done, content in self._prepare_split_single(
File "/usr/local/lib/python3.10/dist-packages/datasets/builder.py", line 2040, in _prepare_split_single
raise DatasetGenerationError("An error occurred while generating the dataset") from e
datasets.exceptions.DatasetGenerationError: An error occurred while generating the dataset
```
```
import pandas as pd
with open("./test.json", "r") as f:
df = pd.read_json(f, dtype_backend="pyarrow")
```
```
Traceback (most recent call last):
File "/app/t3.py", line 3, in <module>
df = pd.read_json(f, dtype_backend="pyarrow")
File "/usr/local/lib/python3.10/dist-packages/pandas/util/_decorators.py", line 211, in wrapper
return func(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/pandas/util/_decorators.py", line 331, in wrapper
return func(*args, **kwargs)
TypeError: read_json() got an unexpected keyword argument 'dtype_backend'
```
### Steps to reproduce the bug
.
### Expected behavior
.
### Environment info
```
datasets 2.20.0
pandas 1.5.3
```
| false
|
2,357,107,203
|
https://api.github.com/repos/huggingface/datasets/issues/6976
|
https://github.com/huggingface/datasets/pull/6976
| 6,976
|
Ensure compatibility with numpy 2.0.0
|
closed
| 2
| 2024-06-17T11:29:22
| 2024-06-19T14:30:32
| 2024-06-19T14:04:34
|
KennethEnevoldsen
|
[] |
Following the conversion guide, copy=False is no longer required and will result in an error: https://numpy.org/devdocs/numpy_2_0_migration_guide.html#adapting-to-changes-in-the-copy-keyword.
The following fix should resolve the issue.
error found during testing on the MTEB repository e.g. [here](https://github.com/embeddings-benchmark/mteb/pull/938)
| true
|
2,357,003,959
|
https://api.github.com/repos/huggingface/datasets/issues/6975
|
https://github.com/huggingface/datasets/pull/6975
| 6,975
|
Set temporary numpy upper version < 2.0.0 to fix CI
|
closed
| 2
| 2024-06-17T10:36:54
| 2024-06-17T12:49:53
| 2024-06-17T12:43:56
|
albertvillanova
|
[] |
Set temporary numpy upper version < 2.0.0 to fix CI. See: https://github.com/huggingface/datasets/actions/runs/9546031216/job/26308072017
```
A module that was compiled using NumPy 1.x cannot be run in
NumPy 2.0.0 as it may crash. To support both 1.x and 2.x
versions of NumPy, modules must be compiled with NumPy 2.0.
Some module may need to rebuild instead e.g. with 'pybind11>=2.12'.
If you are a user of the module, the easiest solution will be to
downgrade to 'numpy<2' or try to upgrade the affected module.
We expect that some modules will need time to support NumPy 2.
```
| true
|
2,355,517,362
|
https://api.github.com/repos/huggingface/datasets/issues/6973
|
https://github.com/huggingface/datasets/issues/6973
| 6,973
|
IndexError during training with Squad dataset and T5-small model
|
closed
| 2
| 2024-06-16T07:53:54
| 2024-07-01T11:25:40
| 2024-07-01T11:25:40
|
ramtunguturi36
|
[] |
### Describe the bug
I am encountering an IndexError while training a T5-small model on the Squad dataset using the transformers and datasets libraries. The error occurs even with a minimal reproducible example, suggesting a potential bug or incompatibility.
### Steps to reproduce the bug
1.Install the required libraries: !pip install transformers datasets
2.Run the following code:
!pip install transformers datasets
import torch
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, TrainingArguments, Trainer, DataCollatorWithPadding
# Load a small, publicly available dataset
from datasets import load_dataset
dataset = load_dataset("squad", split="train[:100]") # Use a small subset for testing
# Load a pre-trained model and tokenizer
model_name = "t5-small"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
# Define a basic data collator
data_collator = DataCollatorWithPadding(tokenizer=tokenizer)
# Define training arguments
training_args = TrainingArguments(
output_dir="./results",
per_device_train_batch_size=2,
num_train_epochs=1,
)
# Create a trainer
trainer = Trainer(
model=model,
args=training_args,
train_dataset=dataset,
data_collator=data_collator,
)
# Train the model
trainer.train()
### Expected behavior
---------------------------------------------------------------------------
IndexError Traceback (most recent call last)
[<ipython-input-23-f13a4b23c001>](https://localhost:8080/#) in <cell line: 34>()
32
33 # Train the model
---> 34 trainer.train()
10 frames
[/usr/local/lib/python3.10/dist-packages/datasets/formatting/formatting.py](https://localhost:8080/#) in _check_valid_index_key(key, size)
427 if isinstance(key, int):
428 if (key < 0 and key + size < 0) or (key >= size):
--> 429 raise IndexError(f"Invalid key: {key} is out of bounds for size {size}")
430 return
431 elif isinstance(key, slice):
IndexError: Invalid key: 42 is out of bounds for size 0
### Environment info
transformers version:4.41.2
datasets version:1.18.4
Python version:3.10.12
| false
|
2,353,531,912
|
https://api.github.com/repos/huggingface/datasets/issues/6972
|
https://github.com/huggingface/datasets/pull/6972
| 6,972
|
Fix webdataset pickling
|
closed
| 2
| 2024-06-14T14:43:02
| 2024-06-14T15:43:43
| 2024-06-14T15:37:35
|
lhoestq
|
[] |
...by making tracked iterables picklable.
This is important to make streaming datasets compatible with multiprocessing e.g. for parallel data loading
| true
|
2,351,830,856
|
https://api.github.com/repos/huggingface/datasets/issues/6971
|
https://github.com/huggingface/datasets/pull/6971
| 6,971
|
packaging: Remove useless dependencies
|
closed
| 4
| 2024-06-13T18:43:43
| 2024-06-14T14:03:34
| 2024-06-14T13:57:24
|
daskol
|
[] |
Revert changes in #6396 and #6404. CVE-2023-47248 has been fixed since PyArrow v14.0.1. Meanwhile Python requirements requires `pyarrow>=15.0.0`.
| true
|
2,351,380,029
|
https://api.github.com/repos/huggingface/datasets/issues/6970
|
https://github.com/huggingface/datasets/pull/6970
| 6,970
|
Set dev version
|
closed
| 2
| 2024-06-13T14:59:45
| 2024-06-13T15:06:18
| 2024-06-13T14:59:56
|
albertvillanova
|
[] | null | true
|
2,351,351,436
|
https://api.github.com/repos/huggingface/datasets/issues/6969
|
https://github.com/huggingface/datasets/pull/6969
| 6,969
|
Release: 2.20.0
|
closed
| 2
| 2024-06-13T14:48:20
| 2024-06-13T15:04:39
| 2024-06-13T14:55:53
|
albertvillanova
|
[] | null | true
|
2,351,331,417
|
https://api.github.com/repos/huggingface/datasets/issues/6968
|
https://github.com/huggingface/datasets/pull/6968
| 6,968
|
Use `HF_HUB_OFFLINE` instead of `HF_DATASETS_OFFLINE`
|
closed
| 3
| 2024-06-13T14:39:40
| 2024-06-13T17:31:37
| 2024-06-13T17:25:37
|
Wauplin
|
[] |
To use `datasets` offline, one can use the `HF_DATASETS_OFFLINE` environment variable. This PR makes `HF_HUB_OFFLINE` the recommended environment variable for offline training. Goal is to be more consistent with the rest of HF ecosystem and have a single config value to set.
The changes are backward-compatible meaning that:
- `HF_DATASETS_OFFLINE` environment is still taken into account, though not documented
- `datasets.config.HF_DATASETS_OFFLINE` still exists, though it is not used anymore (in favor of `datasets.config.HF_HUB_OFFLINE`)
**Note:** it might break things in downstream libraries if they were monkeypatching `datasets.config.HF_DATASETS_OFFLINE` in their CI tests (for instance). Not much of a problem IMO.
| true
|
2,349,146,398
|
https://api.github.com/repos/huggingface/datasets/issues/6967
|
https://github.com/huggingface/datasets/issues/6967
| 6,967
|
Method to load Laion400m
|
open
| 0
| 2024-06-12T16:04:04
| 2024-06-12T16:04:04
| null |
humanely
|
[
"enhancement"
] |
### Feature request
Large datasets like Laion400m are provided as embeddings. The provided methods in load_dataset are not straightforward for loading embedding files, i.e. img_emb_XX.npy ; XX = 0 to 99
### Motivation
The trial and experimentation is the key pivot of HF. It would be great if HF can load embeddings files s,ealessly.
### Your contribution
I cam write the loader with some help.
| false
|
2,348,934,466
|
https://api.github.com/repos/huggingface/datasets/issues/6966
|
https://github.com/huggingface/datasets/pull/6966
| 6,966
|
Remove underlines between badges
|
closed
| 1
| 2024-06-12T14:32:11
| 2024-06-19T14:16:21
| 2024-06-19T14:10:11
|
andrewhong04
|
[] |
## Before:
<img width="935" alt="image" src="https://github.com/huggingface/datasets/assets/35881688/93666e72-059b-4180-9e1d-ff176a3d9dac">
## After:
<img width="956" alt="image" src="https://github.com/huggingface/datasets/assets/35881688/75df7c3e-f473-44f0-a872-eeecf6a85fe2">
| true
|
2,348,653,895
|
https://api.github.com/repos/huggingface/datasets/issues/6965
|
https://github.com/huggingface/datasets/pull/6965
| 6,965
|
Improve skip take shuffling and distributed
|
closed
| 2
| 2024-06-12T12:30:27
| 2024-06-24T15:22:21
| 2024-06-24T15:16:16
|
lhoestq
|
[] |
set the right behavior of skip/take depending on whether it's called after or before shuffle/split_by_node
| true
|
2,344,973,229
|
https://api.github.com/repos/huggingface/datasets/issues/6964
|
https://github.com/huggingface/datasets/pull/6964
| 6,964
|
Fix resuming arrow format
|
closed
| 2
| 2024-06-10T22:40:33
| 2024-06-14T15:04:49
| 2024-06-14T14:58:37
|
lhoestq
|
[] |
following https://github.com/huggingface/datasets/pull/6658
| true
|
2,344,269,477
|
https://api.github.com/repos/huggingface/datasets/issues/6963
|
https://github.com/huggingface/datasets/pull/6963
| 6,963
|
[Streaming] retry on requests errors
|
closed
| 3
| 2024-06-10T15:51:56
| 2024-06-28T09:53:11
| 2024-06-28T09:46:52
|
lhoestq
|
[] |
reported in https://discuss.huggingface.co/t/speeding-up-streaming-of-large-datasets-fineweb/90714/6 when training using a streaming a dataloader
cc @Wauplin it looks like the retries from `hfh` are not always enough. In this PR I let `datasets` do additional retries (that users can configure in `datasets.config`) since I couldn't find an easy way to increase the max_retries for `hfh` users in general.
| true
|
2,343,394,378
|
https://api.github.com/repos/huggingface/datasets/issues/6962
|
https://github.com/huggingface/datasets/pull/6962
| 6,962
|
fix(ci): remove unnecessary permissions
|
closed
| 2
| 2024-06-10T09:28:02
| 2024-06-11T08:31:52
| 2024-06-11T08:25:47
|
McPatate
|
[] |
### What does this PR do?
Remove unnecessary permissions granted to the actions workflow.
Sorry for the mishap.
| true
|
2,342,022,418
|
https://api.github.com/repos/huggingface/datasets/issues/6961
|
https://github.com/huggingface/datasets/issues/6961
| 6,961
|
Manual downloads should count as downloads
|
open
| 1
| 2024-06-09T04:52:06
| 2024-06-13T16:05:00
| null |
umarbutler
|
[
"enhancement"
] |
### Feature request
I would like to request that manual downloads of data files from Hugging Face dataset repositories count as downloads of a dataset. According to the documentation for the Hugging Face Hub, that is currently not the case: https://huggingface.co/docs/hub/en/datasets-download-stats
### Motivation
This would ensure that downloads are accurately reported to end users.
### Your contribution
N/A
| false
|
2,340,791,685
|
https://api.github.com/repos/huggingface/datasets/issues/6960
|
https://github.com/huggingface/datasets/pull/6960
| 6,960
|
feat(ci): add trufflehog secrets detection
|
closed
| 3
| 2024-06-07T16:18:23
| 2024-06-08T14:58:27
| 2024-06-08T14:52:18
|
McPatate
|
[] |
### What does this PR do?
Adding a GH action to scan for leaked secrets on each commit.
| true
|
2,340,229,908
|
https://api.github.com/repos/huggingface/datasets/issues/6959
|
https://github.com/huggingface/datasets/pull/6959
| 6,959
|
Better error handling in `dataset_module_factory`
|
closed
| 3
| 2024-06-07T11:24:15
| 2024-06-10T07:33:53
| 2024-06-10T07:27:43
|
Wauplin
|
[] |
cc @cakiki who reported it on [slack](https://huggingface.slack.com/archives/C039P47V1L5/p1717754405578539) (private link)
This PR updates how errors are handled in `dataset_module_factory` when the `dataset_info` cannot be accessed:
1. Use multiple `except ... as e` instead of using `isinstance(e, ...)`
2. Always raise `DatasetNotFoundError` with `from e` so that the initial error is explicitly logged in the stacktrace.
3. Differentiate `RepoNotFoundError` / `GatedRepoError` / `RevisionNotFoundError` cases
| true
|
2,337,476,383
|
https://api.github.com/repos/huggingface/datasets/issues/6958
|
https://github.com/huggingface/datasets/issues/6958
| 6,958
|
My Private Dataset doesn't exist on the Hub or cannot be accessed
|
closed
| 8
| 2024-06-06T06:52:19
| 2024-07-01T11:27:46
| 2024-07-01T11:27:46
|
wangguan1995
|
[] |
### Describe the bug
```
File "/root/miniconda3/envs/gino_conda/lib/python3.9/site-packages/datasets/load.py", line 1852, in dataset_module_factory
raise DatasetNotFoundError(msg + f" at revision '{revision}'" if revision else msg)
datasets.exceptions.DatasetNotFoundError: Dataset 'xxx' doesn't exist on the Hub or cannot be accessed
>>> dataset = load_dataset("xxxx", token=True)
404 error 404 Client Error. (Request ID: Root=xxxx)
Repository Not Found for url: https://huggingface.co/api/datasets/xxx/xxx.
Please make sure you specified the correct `repo_id` and `repo_type`.
If you are trying to access a private or gated repo, make sure you are authenticated.
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/root/miniconda3/envs/gino_conda/lib/python3.9/site-packages/datasets/load.py", line 2593, in load_dataset
builder_instance = load_dataset_builder(
File "/root/miniconda3/envs/gino_conda/lib/python3.9/site-packages/datasets/load.py", line 2265, in load_dataset_builder
dataset_module = dataset_module_factory(
File "/root/miniconda3/envs/gino_conda/lib/python3.9/site-packages/datasets/load.py", line 1910, in dataset_module_factory
raise e1 from None
File "/root/miniconda3/envs/gino_conda/lib/python3.9/site-packages/datasets/load.py", line 1852, in dataset_module_factory
raise DatasetNotFoundError(msg + f" at revision '{revision}'" if revision else msg)
datasets.exceptions.DatasetNotFoundError: Dataset 'xxx' doesn't exist on the Hub or cannot be accessed
```
### Steps to reproduce the bug
123
### Expected behavior
123
### Environment info
123
| false
|
2,335,559,400
|
https://api.github.com/repos/huggingface/datasets/issues/6957
|
https://github.com/huggingface/datasets/pull/6957
| 6,957
|
Fix typos in docs
|
closed
| 2
| 2024-06-05T10:46:47
| 2024-06-05T13:01:07
| 2024-06-05T12:43:26
|
albertvillanova
|
[] |
Fix typos in docs introduced by:
- #6956
Typos:
- `comparisions` => `comparisons`
- two consecutive sentences both ending in colon
- split one sentence into two
Sorry, I did not have time to review that PR.
CC: @lhoestq
| true
|
2,333,940,021
|
https://api.github.com/repos/huggingface/datasets/issues/6956
|
https://github.com/huggingface/datasets/pull/6956
| 6,956
|
update docs on N-dim arrays
|
closed
| 2
| 2024-06-04T16:32:19
| 2024-06-04T16:46:34
| 2024-06-04T16:40:27
|
lhoestq
|
[] | null | true
|
2,333,802,815
|
https://api.github.com/repos/huggingface/datasets/issues/6955
|
https://github.com/huggingface/datasets/pull/6955
| 6,955
|
Fix small typo
|
closed
| 1
| 2024-06-04T15:19:02
| 2024-06-05T10:18:56
| 2024-06-04T15:20:55
|
marcenacp
|
[] | null | true
|
2,333,530,558
|
https://api.github.com/repos/huggingface/datasets/issues/6954
|
https://github.com/huggingface/datasets/pull/6954
| 6,954
|
Remove default `trust_remote_code=True`
|
closed
| 6
| 2024-06-04T13:22:56
| 2024-06-17T16:32:24
| 2024-06-07T12:20:29
|
lhoestq
|
[] |
TODO:
- [x] fix tests
| true
|
2,333,366,120
|
https://api.github.com/repos/huggingface/datasets/issues/6953
|
https://github.com/huggingface/datasets/issues/6953
| 6,953
|
Remove canonical datasets from docs
|
closed
| 1
| 2024-06-04T12:09:03
| 2024-07-01T11:31:25
| 2024-07-01T11:31:25
|
albertvillanova
|
[
"documentation"
] |
Remove canonical datasets from docs, now that we no longer have canonical datasets.
| false
|
2,333,320,411
|
https://api.github.com/repos/huggingface/datasets/issues/6952
|
https://github.com/huggingface/datasets/pull/6952
| 6,952
|
Move info_utils errors to exceptions module
|
closed
| 2
| 2024-06-04T11:48:32
| 2024-06-10T14:09:59
| 2024-06-10T14:03:55
|
albertvillanova
|
[] |
Move `info_utils` errors to `exceptions` module.
Additionally rename some of them, deprecate the former ones, and make the deprecation backward compatible (by making the new errors inherit from the former ones).
| true
|
2,333,231,042
|
https://api.github.com/repos/huggingface/datasets/issues/6951
|
https://github.com/huggingface/datasets/issues/6951
| 6,951
|
load_dataset() should load all subsets, if no specific subset is specified
|
closed
| 5
| 2024-06-04T11:02:33
| 2024-11-26T08:32:18
| 2024-07-01T11:33:10
|
windmaple
|
[
"enhancement"
] |
### Feature request
Currently load_dataset() is forcing users to specify a subset. Example
`from datasets import load_dataset
dataset = load_dataset("m-a-p/COIG-CQIA")`
```---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
[<ipython-input-10-c0cb49385da6>](https://localhost:8080/#) in <cell line: 2>()
1 from datasets import load_dataset
----> 2 dataset = load_dataset("m-a-p/COIG-CQIA")
3 frames
[/usr/local/lib/python3.10/dist-packages/datasets/builder.py](https://localhost:8080/#) in _create_builder_config(self, config_name, custom_features, **config_kwargs)
582 if not config_kwargs:
583 example_of_usage = f"load_dataset('{self.dataset_name}', '{self.BUILDER_CONFIGS[0].name}')"
--> 584 raise ValueError(
585 "Config name is missing."
586 f"\nPlease pick one among the available configs: {list(self.builder_configs.keys())}"
ValueError: Config name is missing.
Please pick one among the available configs: ['chinese_traditional', 'coig_pc', 'exam', 'finance', 'douban', 'human_value', 'logi_qa', 'ruozhiba', 'segmentfault', 'wiki', 'wikihow', 'xhs', 'zhihu']
Example of usage:
`load_dataset('coig-cqia', 'chinese_traditional')`
```
This means a dataset cannot contain all the subsets at the same time. I guess one workaround is to manually specify the subset files like in [here](https://huggingface.co/datasets/m-a-p/COIG-CQIA/discussions/1#658698b44bb41498f75c5622), which is clumsy.
### Motivation
Ideally, if not subset is specified, the API should just try to load all subsets. This makes it much easier to handle datasets w/ subsets.
### Your contribution
Not sure since I'm not familiar w/ the lib src.
| false
|
2,333,005,974
|
https://api.github.com/repos/huggingface/datasets/issues/6950
|
https://github.com/huggingface/datasets/issues/6950
| 6,950
|
`Dataset.with_format` behaves inconsistently with documentation
|
closed
| 2
| 2024-06-04T09:18:32
| 2024-06-25T08:05:49
| 2024-06-25T08:05:49
|
iansheng
|
[
"documentation"
] |
### Describe the bug
The actual behavior of the interface `Dataset.with_format` is inconsistent with the documentation.
https://huggingface.co/docs/datasets/use_with_pytorch#n-dimensional-arrays
https://huggingface.co/docs/datasets/v2.19.0/en/use_with_tensorflow#n-dimensional-arrays
> If your dataset consists of N-dimensional arrays, you will see that by default they are considered as nested lists.
> In particular, a PyTorch formatted dataset outputs nested lists instead of a single tensor.
> A TensorFlow formatted dataset outputs a RaggedTensor instead of a single tensor.
But I get a single tensor by default, which is inconsistent with the description.
Actually the current behavior seems more reasonable to me. Therefore, the document needs to be modified.
### Steps to reproduce the bug
```python
>>> from datasets import Dataset
>>> data = [[[1, 2],[3, 4]],[[5, 6],[7, 8]]]
>>> ds = Dataset.from_dict({"data": data})
>>> ds = ds.with_format("torch")
>>> ds[0]
{'data': tensor([[1, 2],
[3, 4]])}
>>> ds = ds.with_format("tf")
>>> ds[0]
{'data': <tf.Tensor: shape=(2, 2), dtype=int64, numpy=
array([[1, 2],
[3, 4]])>}
```
### Expected behavior
```python
>>> from datasets import Dataset
>>> data = [[[1, 2],[3, 4]],[[5, 6],[7, 8]]]
>>> ds = Dataset.from_dict({"data": data})
>>> ds = ds.with_format("torch")
>>> ds[0]
{'data': [tensor([1, 2]), tensor([3, 4])]}
>>> ds = ds.with_format("tf")
>>> ds[0]
{'data': <tf.RaggedTensor [[1, 2], [3, 4]]>}
```
### Environment info
datasets==2.19.1
torch==2.1.0
tensorflow==2.13.1
| false
|
2,332,336,573
|
https://api.github.com/repos/huggingface/datasets/issues/6949
|
https://github.com/huggingface/datasets/issues/6949
| 6,949
|
load_dataset error
|
closed
| 2
| 2024-06-04T01:24:45
| 2024-07-01T11:33:46
| 2024-07-01T11:33:46
|
frederichen01
|
[] |
### Describe the bug
Why does the program get stuck when I use load_dataset method, and it still gets stuck after loading for several hours? In fact, my json file is only 21m, and I can load it in one go using open('', 'r').
### Steps to reproduce the bug
1. pip install datasets==2.19.2
2. from datasets import Dataset, DatasetDict, NamedSplit, Split, load_dataset
3. data = load_dataset('json', data_files='train.json')
### Expected behavior
It is able to load my json correctly
### Environment info
datasets==2.19.2
| false
|
2,331,758,300
|
https://api.github.com/repos/huggingface/datasets/issues/6948
|
https://github.com/huggingface/datasets/issues/6948
| 6,948
|
to_tf_dataset: Visible devices cannot be modified after being initialized
|
open
| 0
| 2024-06-03T18:10:57
| 2024-06-03T18:10:57
| null |
logasja
|
[] |
### Describe the bug
When trying to use to_tf_dataset with a custom data_loader collate_fn when I use parallelism I am met with the following error as many times as number of workers there were in ``num_workers``.
File "/opt/miniconda/envs/env/lib/python3.11/site-packages/multiprocess/process.py", line 314, in _bootstrap
self.run()
File "/opt/miniconda/envs/env/lib/python3.11/site-packages/multiprocess/process.py", line 108, in run
self._target(*self._args, **self._kwargs)
File "/opt/miniconda/envs/env/lib/python3.11/site-packages/datasets/utils/tf_utils.py", line 438, in worker_loop
tf.config.set_visible_devices([], "GPU") # Make sure workers don't try to allocate GPU memory
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/miniconda/envs/env/lib/python3.11/site-packages/tensorflow/python/framework/config.py", line 566, in set_visible_devices
context.context().set_visible_devices(devices, device_type)
File "/opt/miniconda/envs/env/lib/python3.11/site-packages/tensorflow/python/eager/context.py", line 1737, in set_visible_devices
raise RuntimeError(
RuntimeError: Visible devices cannot be modified after being initialized
### Steps to reproduce the bug
1. Download a dataset using HuggingFace load_dataset
2. Define a function that transforms the data in some way to be used in the collate_fn argument
3. Provide a ``batch_size`` and ``num_workers`` value in the ``to_tf_dataset`` function
4. Either retrieve directly or use tfds benchmark to test the dataset
``` python
from datasets import load_datasets
import tensorflow_datasets as tfds
from keras_cv.layers import Resizing
def data_loader(examples):
x = Resizing(examples[0]['image'], 256, 256, crop_to_aspect_ratio=True)
return {X[0]: x}
ds = load_datasets("logasja/FDF", split="test")
ds = ds.to_tf_dataset(collate_fn=data_loader, batch_size=16, num_workers=2)
tfds.benchmark(ds)
```
### Expected behavior
Use multiple processes to apply transformations from the collate_fn to the tf dataset on the CPU.
### Environment info
- `datasets` version: 2.19.1
- Platform: Linux-6.5.0-1023-oracle-x86_64-with-glibc2.35
- Python version: 3.11.8
- `huggingface_hub` version: 0.22.2
- PyArrow version: 15.0.2
- Pandas version: 2.2.1
- `fsspec` version: 2024.2.0
| false
|
2,331,114,055
|
https://api.github.com/repos/huggingface/datasets/issues/6947
|
https://github.com/huggingface/datasets/issues/6947
| 6,947
|
FileNotFoundError：error when loading C4 dataset
|
closed
| 15
| 2024-06-03T13:06:33
| 2024-06-25T06:21:28
| 2024-06-25T06:21:28
|
W-215
|
[] |
### Describe the bug
can't load c4 datasets
When I replace the datasets package to 2.12.2 I get raise datasets.utils.info_utils.ExpectedMoreSplits: {'train'}
How can I fix this？
### Steps to reproduce the bug
1.from datasets import load_dataset
2.dataset = load_dataset('allenai/c4', data_files={'validation': 'en/c4-validation.00003-of-00008.json.gz'}, split='validation')
3. raise FileNotFoundError(
FileNotFoundError: Couldn't find a dataset script at local_path/c4_val/allenai/c4/c4.py or any data file in the same directory. Couldn't find 'allenai/c4' on the Hugging Face Hub either: FileNotFoundError: Unable to find 'hf://datasets/allenai/c4@1588ec454efa1a09f29cd18ddd04fe05fc8653a2/en/c4-validation.00003-of-00008.json.gz' with any supported extension ['.csv', '.tsv', '.json', '.jsonl', '.parquet', '.geoparquet', '.gpq', '.arrow', '.txt', '.tar', '.blp', '.bmp', '.dib', '.bufr', '.cur', '.pcx', '.dcx', '.dds', '.ps', '.eps', '.fit', '.fits', '.fli', '.flc', '.ftc', '.ftu', '.gbr', '.gif', '.grib', '.h5', '.hdf', '.png', '.apng', '.jp2', '.j2k', '.jpc', '.jpf', '.jpx', '.j2c', '.icns', '.ico', '.im', '.iim', '.tif', '.tiff', '.jfif', '.jpe', '.jpg', '.jpeg', '.mpg', '.mpeg', '.msp', '.pcd', '.pxr', '.pbm', '.pgm', '.ppm', '.pnm', '.psd', '.bw', '.rgb', '.rgba', '.sgi', '.ras', '.tga', '.icb', '.vda', '.vst', '.webp', '.wmf', '.emf', '.xbm', '.xpm', '.BLP', '.BMP', '.DIB', '.BUFR', '.CUR', '.PCX', '.DCX', '.DDS', '.PS', '.EPS', '.FIT', '.FITS', '.FLI', '.FLC', '.FTC', '.FTU', '.GBR', '.GIF', '.GRIB', '.H5', '.HDF', '.PNG', '.APNG', '.JP2', '.J2K', '.JPC', '.JPF', '.JPX', '.J2C', '.ICNS', '.ICO', '.IM', '.IIM', '.TIF', '.TIFF', '.JFIF', '.JPE', '.JPG', '.JPEG', '.MPG', '.MPEG', '.MSP', '.PCD', '.PXR', '.PBM', '.PGM', '.PPM', '.PNM', '.PSD', '.BW', '.RGB', '.RGBA', '.SGI', '.RAS', '.TGA', '.ICB', '.VDA', '.VST', '.WEBP', '.WMF', '.EMF', '.XBM', '.XPM', '.aiff', '.au', '.avr', '.caf', '.flac', '.htk', '.svx', '.mat4', '.mat5', '.mpc2k', '.ogg', '.paf', '.pvf', '.raw', '.rf64', '.sd2', '.sds', '.ircam', '.voc', '.w64', '.wav', '.nist', '.wavex', '.wve', '.xi', '.mp3', '.opus', '.AIFF', '.AU', '.AVR', '.CAF', '.FLAC', '.HTK', '.SVX', '.MAT4', '.MAT5', '.MPC2K', '.OGG', '.PAF', '.PVF', '.RAW', '.RF64', '.SD2', '.SDS', '.IRCAM', '.VOC', '.W64', '.WAV', '.NIST', '.WAVEX', '.WVE', '.XI', '.MP3', '.OPUS', '.zip']
### Expected behavior
The data was successfully imported
### Environment info
python version 3.9
datasets version 2.19.2
| false
|
2,330,276,848
|
https://api.github.com/repos/huggingface/datasets/issues/6946
|
https://github.com/huggingface/datasets/pull/6946
| 6,946
|
Re-enable import sorting disabled by flake8:noqa directive when using ruff linter
|
closed
| 2
| 2024-06-03T06:24:47
| 2024-06-04T10:00:08
| 2024-06-04T09:54:23
|
albertvillanova
|
[] |
Re-enable import sorting that was wrongly disabled by `flake8: noqa` directive after switching to `ruff` linter in datasets-2.10.0 PR:
- #5519
Note that after the linter switch, we wrongly replaced `flake8: noqa` with `ruff: noqa` in datasets-2.17.0 PR:
- #6619
That replacement was wrong because we kept the `isort: skip` directives although they were indeed disabled by `flake8: noqa` first and by `ruff: noqa` afterwards. See for example `__init__.py` file after the linter switch:
- We kept the `flake8: noqa` directive
https://github.com/huggingface/datasets/blob/06ae3f678651bfbb3ca7dd3274ee2f38e0e0237e/src/datasets/__init__.py#L1
- Whereas we also kept the `isort: skip` directives (that were disabled)
https://github.com/huggingface/datasets/blob/06ae3f678651bfbb3ca7dd3274ee2f38e0e0237e/src/datasets/__init__.py#L82-L84
Fix #6942.
| true
|
2,330,224,869
|
https://api.github.com/repos/huggingface/datasets/issues/6945
|
https://github.com/huggingface/datasets/pull/6945
| 6,945
|
Update yanked version of minimum requests requirement
|
closed
| 5
| 2024-06-03T05:45:50
| 2024-06-18T07:36:15
| 2024-06-03T06:09:43
|
albertvillanova
|
[] |
Update yanked version of minimum requests requirement.
Version 2.32.1 was yanked: https://pypi.org/project/requests/2.32.1/
| true
|
2,330,207,120
|
https://api.github.com/repos/huggingface/datasets/issues/6944
|
https://github.com/huggingface/datasets/pull/6944
| 6,944
|
Set dev version
|
closed
| 2
| 2024-06-03T05:29:59
| 2024-06-03T05:37:51
| 2024-06-03T05:31:47
|
albertvillanova
|
[] | null | true
|
2,330,176,890
|
https://api.github.com/repos/huggingface/datasets/issues/6943
|
https://github.com/huggingface/datasets/pull/6943
| 6,943
|
Release 2.19.2
|
closed
| 1
| 2024-06-03T05:01:50
| 2024-06-03T05:17:41
| 2024-06-03T05:17:40
|
albertvillanova
|
[] | null | true
|
2,329,562,382
|
https://api.github.com/repos/huggingface/datasets/issues/6942
|
https://github.com/huggingface/datasets/issues/6942
| 6,942
|
Import sorting is disabled by flake8 noqa directive after switching to ruff linter
|
closed
| 0
| 2024-06-02T09:43:34
| 2024-06-04T09:54:24
| 2024-06-04T09:54:24
|
albertvillanova
|
[
"maintenance"
] |
When we switched to `ruff` linter in PR:
- #5519
import sorting was disabled in all files containing the `# flake8: noqa` directive
- https://github.com/astral-sh/ruff/issues/11679
We should re-enable import sorting on those files.
| false
|
2,328,930,165
|
https://api.github.com/repos/huggingface/datasets/issues/6941
|
https://github.com/huggingface/datasets/issues/6941
| 6,941
|
Supporting FFCV: Fast Forward Computer Vision
|
open
| 0
| 2024-06-01T05:34:52
| 2024-06-01T05:34:52
| null |
Luciennnnnnn
|
[
"enhancement"
] |
### Feature request
Supporting FFCV, https://github.com/libffcv/ffcv
### Motivation
According to the benchmark, FFCV seems to be fastest image loading method.
### Your contribution
no
| false
|
2,328,637,831
|
https://api.github.com/repos/huggingface/datasets/issues/6940
|
https://github.com/huggingface/datasets/issues/6940
| 6,940
|
Enable Sharding to Equal Sized Shards
|
open
| 0
| 2024-05-31T21:55:50
| 2024-06-01T07:34:12
| null |
yuvalkirstain
|
[
"enhancement"
] |
### Feature request
Add an option when sharding a dataset to have all shards the same size. Will be good to provide both an option of duplication, and by truncation.
### Motivation
Currently the behavior of sharding is "If n % i == l, then the first l shards will have length (n // i) + 1, and the remaining shards will have length (n // i).". However, when using FSDP we want the shards to have the same size. This requires the user to manually handle this situation, but it will be nice if we had an option to shard the dataset into equally sized shards.
### Your contribution
For now just a PR. I can also add code that does what is needed, but probably not efficient.
Shard to equal size by duplication:
```
remainder = len(dataset) % num_shards
num_missing_examples = num_shards - remainder
duplicated = dataset.select(list(range(num_missing_examples)))
dataset = concatenate_datasets([dataset, duplicated])
shard = dataset.shard(num_shards, shard_idx)
```
Or by truncation:
```
shard = dataset.shard(num_shards, shard_idx)
num_examples_per_shard = len(dataset) // num_shards
shard = shard.select(list(range(num_examples_per_shard)))
```
| false
|
2,328,059,386
|
https://api.github.com/repos/huggingface/datasets/issues/6939
|
https://github.com/huggingface/datasets/issues/6939
| 6,939
|
ExpectedMoreSplits error when using data_dir
|
closed
| 0
| 2024-05-31T15:08:42
| 2024-05-31T17:10:39
| 2024-05-31T17:10:39
|
albertvillanova
|
[
"bug"
] |
As reported by @regisss, an `ExpectedMoreSplits` error is raised when passing `data_dir`:
```python
from datasets import load_dataset
dataset = load_dataset(
"lvwerra/stack-exchange-paired",
split="train",
cache_dir=None,
data_dir="data/rl",
)
```
```
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/usr/local/lib/python3.10/dist-packages/datasets/load.py", line 2609, in load_dataset
builder_instance.download_and_prepare(
File "/usr/local/lib/python3.10/dist-packages/datasets/builder.py", line 1027, in download_and_prepare
self._download_and_prepare(
File "/usr/local/lib/python3.10/dist-packages/datasets/builder.py", line 1140, in _download_and_prepare
verify_splits(self.info.splits, split_dict)
File "/usr/local/lib/python3.10/dist-packages/datasets/utils/info_utils.py", line 92, in verify_splits
raise ExpectedMoreSplits(str(set(expected_splits) - set(recorded_splits)))
datasets.utils.info_utils.ExpectedMoreSplits: {'test'}
```
| false
|
2,327,568,281
|
https://api.github.com/repos/huggingface/datasets/issues/6938
|
https://github.com/huggingface/datasets/pull/6938
| 6,938
|
Fix expected splits when passing data_files or dir
|
closed
| 2
| 2024-05-31T11:04:22
| 2024-05-31T15:28:03
| 2024-05-31T15:28:02
|
lhoestq
|
[] |
reported on slack:
The following code snippet gives an error with v2.19 but not with v2.18:
from datasets import load_dataset
```
dataset = load_dataset(
"lvwerra/stack-exchange-paired",
split="train",
cache_dir=None,
data_dir="data/rl",
)
```
and the error is:
```
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/usr/local/lib/python3.10/dist-packages/datasets/load.py", line 2609, in load_dataset
builder_instance.download_and_prepare(
File "/usr/local/lib/python3.10/dist-packages/datasets/builder.py", line 1027, in download_and_prepare
self._download_and_prepare(
File "/usr/local/lib/python3.10/dist-packages/datasets/builder.py", line 1140, in _download_and_prepare
verify_splits(self.info.splits, split_dict)
File "/usr/local/lib/python3.10/dist-packages/datasets/utils/info_utils.py", line 92, in verify_splits
raise ExpectedMoreSplits(str(set(expected_splits) - set(recorded_splits)))
datasets.utils.info_utils.ExpectedMoreSplits: {'test'}
```
| true
|
2,327,212,611
|
https://api.github.com/repos/huggingface/datasets/issues/6937
|
https://github.com/huggingface/datasets/issues/6937
| 6,937
|
JSON loader implicitly coerces floats to integers
|
open
| 1
| 2024-05-31T08:09:12
| 2025-06-24T05:49:20
| null |
albertvillanova
|
[
"bug"
] |
The JSON loader implicitly coerces floats to integers.
The column values `[0.0, 1.0, 2.0]` are coerced to `[0, 1, 2]`.
See CI error in dataset-viewer: https://github.com/huggingface/dataset-viewer/actions/runs/9290164936/job/25576926446
```
=================================== FAILURES ===================================
___________________________ test_statistics_endpoint ___________________________
normal_user_public_json_dataset = 'DVUser/tmp-dataset-17170199043860'
def test_statistics_endpoint(normal_user_public_json_dataset: str) -> None:
dataset = normal_user_public_json_dataset
config, split = get_default_config_split()
statistics_response = poll_until_ready_and_assert(
relative_url=f"/statistics?dataset={dataset}&config={config}&split={split}",
check_x_revision=True,
dataset=dataset,
)
content = statistics_response.json()
assert len(content) == 3
assert sorted(content) == ["num_examples", "partial", "statistics"], statistics_response
statistics = content["statistics"]
num_examples = content["num_examples"]
partial = content["partial"]
assert isinstance(statistics, list), statistics
assert len(statistics) == 6
assert num_examples == 4
assert partial is False
string_label_column = statistics[0]
assert "column_name" in string_label_column
assert "column_statistics" in string_label_column
assert "column_type" in string_label_column
assert string_label_column["column_name"] == "col_1"
assert string_label_column["column_type"] == "string_label" # 4 unique values -> label
assert isinstance(string_label_column["column_statistics"], dict)
assert string_label_column["column_statistics"] == {
"nan_count": 0,
"nan_proportion": 0.0,
"no_label_count": 0,
"no_label_proportion": 0.0,
"n_unique": 4,
"frequencies": {
"There goes another one.": 1,
"Vader turns round and round in circles as his ship spins into space.": 1,
"We count thirty Rebel ships, Lord Vader.": 1,
"The wingman spots the pirateship coming at him and warns the Dark Lord": 1,
},
}
int_column = statistics[1]
assert "column_name" in int_column
assert "column_statistics" in int_column
assert "column_type" in int_column
assert int_column["column_name"] == "col_2"
assert int_column["column_type"] == "int"
assert isinstance(int_column["column_statistics"], dict)
assert int_column["column_statistics"] == {
"histogram": {"bin_edges": [0, 1, 2, 3, 3], "hist": [1, 1, 1, 1]},
"max": 3,
"mean": 1.5,
"median": 1.5,
"min": 0,
"nan_count": 0,
"nan_proportion": 0.0,
"std": 1.29099,
}
float_column = statistics[2]
assert "column_name" in float_column
assert "column_statistics" in float_column
assert "column_type" in float_column
assert float_column["column_name"] == "col_3"
> assert float_column["column_type"] == "float"
E AssertionError: assert 'int' == 'float'
E - float
E + int
tests/test_14_statistics.py:72: AssertionError
=========================== short test summary info ============================
FAILED tests/test_14_statistics.py::test_statistics_endpoint - AssertionError: assert 'int' == 'float'
- float
+ int
```
This bug was introduced after:
- #6914
We have reported the issue to pandas:
- https://github.com/pandas-dev/pandas/issues/58866
| false
|
2,326,119,853
|
https://api.github.com/repos/huggingface/datasets/issues/6936
|
https://github.com/huggingface/datasets/issues/6936
| 6,936
|
save_to_disk() freezes when saving on s3 bucket with multiprocessing
|
open
| 3
| 2024-05-30T16:48:39
| 2025-02-06T22:12:52
| null |
ycattan
|
[] |
### Describe the bug
I'm trying to save a `Dataset` using the `save_to_disk()` function with:
- `num_proc > 1`
- `dataset_path` being a s3 bucket path e.g. "s3://{bucket_name}/{dataset_folder}/"
The hf progress bar shows up but the saving does not seem to start.
When using one processor only (`num_proc=1`), everything works fine.
When saving the dataset on local disk (as opposed to s3 bucket) with `num_proc > 1`, everything works fine.
Thank you for your help! :)
### Steps to reproduce the bug
I tried without any storage options:
```
from datasets import load_dataset
sandbox_ds = load_dataset("openai_humaneval")
sandbox_ds["test"].save_to_disk(
"s3://bucket-name/test_multiprocessing_saving/",
num_proc=4,
)
```
and with the specific s3fs storage options:
```
from datasets import load_dataset
from s3fs import S3FileSystem
def get_s3fs():
return S3FileSystem()
sandbox_ds = load_dataset("openai_humaneval")
sandbox_ds["test"].save_to_disk(
"s3://bucket-name/test_multiprocessing_saving/",
num_proc=4,
storage_options=get_s3fs().storage_options, # also tried: storage_options=S3FileSystem().storage_options
)
```
I'm guessing I might use `storage_options` parameter wrongly, but I didn't find anything online that made it work.
**NB**: Behavior is the same when trying to save the whole `DatasetDict`.
### Expected behavior
Progress bar fills in and saving is carried out.
### Environment info
`datasets==2.18.0`
| false
|
2,325,612,022
|
https://api.github.com/repos/huggingface/datasets/issues/6935
|
https://github.com/huggingface/datasets/issues/6935
| 6,935
|
Support for pathlib.Path in datasets 2.19.0
|
open
| 2
| 2024-05-30T12:53:36
| 2025-01-14T11:50:22
| null |
lamyiowce
|
[] |
### Describe the bug
After the recent update of `datasets`, Dataset.save_to_disk does not accept a pathlib.Path anymore. It was supported in 2.18.0 and previous versions. Is this intentional? Was it supported before only because of a Python dusk-typing miracle?
### Steps to reproduce the bug
```
from datasets import Dataset
import pathlib
path = pathlib.Path("./my_out_path")
Dataset.from_dict(
{"text": ["hello world"], "label": [777], "split": ["train"]}
.save_to_disk(path)
```
This results in an error when using datasets 2.19:
```
Traceback (most recent call last):
File "<stdin>", line 3, in <module>
File "/Users/jb/scratch/venv/lib/python3.11/site-packages/datasets/arrow_dataset.py", line 1515, in save_to_disk
fs, _ = url_to_fs(dataset_path, **(storage_options or {}))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/jb/scratch/venv/lib/python3.11/site-packages/fsspec/core.py", line 383, in url_to_fs
chain = _un_chain(url, kwargs)
^^^^^^^^^^^^^^^^^^^^^^
File "/Users/jb/scratch/venv/lib/python3.11/site-packages/fsspec/core.py", line 323, in _un_chain
if "::" in path
^^^^^^^^^^^^
TypeError: argument of type 'PosixPath' is not iterable
```
Converting to str works, however.
```
Dataset.from_dict(
{"text": ["hello world"], "label": [777], "split": ["train"]}
).save_to_disk(str(path))
```
### Expected behavior
My dataset gets saved to disk without an error.
### Environment info
aiohttp==3.9.5
aiosignal==1.3.1
attrs==23.2.0
certifi==2024.2.2
charset-normalizer==3.3.2
datasets==2.19.0
dill==0.3.8
filelock==3.14.0
frozenlist==1.4.1
fsspec==2024.3.1
huggingface-hub==0.23.2
idna==3.7
multidict==6.0.5
multiprocess==0.70.16
numpy==1.26.4
packaging==24.0
pandas==2.2.2
pyarrow==16.1.0
pyarrow-hotfix==0.6
python-dateutil==2.9.0.post0
pytz==2024.1
PyYAML==6.0.1
requests==2.32.3
six==1.16.0
tqdm==4.66.4
typing_extensions==4.12.0
tzdata==2024.1
urllib3==2.2.1
xxhash==3.4.1
yarl==1.9.4
| false
|
2,325,341,717
|
https://api.github.com/repos/huggingface/datasets/issues/6934
|
https://github.com/huggingface/datasets/pull/6934
| 6,934
|
Revert ci user
|
closed
| 3
| 2024-05-30T10:45:26
| 2024-05-31T10:25:08
| 2024-05-30T10:45:37
|
lhoestq
|
[] | null | true
|
2,325,300,800
|
https://api.github.com/repos/huggingface/datasets/issues/6933
|
https://github.com/huggingface/datasets/pull/6933
| 6,933
|
update ci user
|
closed
| 2
| 2024-05-30T10:23:02
| 2024-05-30T10:30:54
| 2024-05-30T10:23:12
|
lhoestq
|
[] |
token is ok to be public since it's only for the hub-ci
| true
|
2,324,729,267
|
https://api.github.com/repos/huggingface/datasets/issues/6932
|
https://github.com/huggingface/datasets/pull/6932
| 6,932
|
Update dataset_dict.py
|
closed
| 2
| 2024-05-30T05:22:35
| 2024-06-04T12:56:20
| 2024-06-04T12:50:13
|
Arunprakash-A
|
[] |
shape returns (number of rows, number of columns)
| true
|
2,323,457,525
|
https://api.github.com/repos/huggingface/datasets/issues/6931
|
https://github.com/huggingface/datasets/pull/6931
| 6,931
|
[WebDataset] Support compressed files
|
closed
| 2
| 2024-05-29T14:19:06
| 2024-05-29T16:33:18
| 2024-05-29T16:24:21
|
lhoestq
|
[] | null | true
|
2,323,225,922
|
https://api.github.com/repos/huggingface/datasets/issues/6930
|
https://github.com/huggingface/datasets/issues/6930
| 6,930
|
ValueError: Couldn't infer the same data file format for all splits. Got {'train': ('json', {}), 'validation': (None, {})}
|
open
| 2
| 2024-05-29T12:40:05
| 2024-07-23T06:25:24
| null |
Polarisamoon
|
[] |
### Describe the bug
When I run the code en = load_dataset("allenai/c4", "en", streaming=True), I encounter an error: raise ValueError(f"Couldn't infer the same data file format for all splits. Got {split_modules}") ValueError: Couldn't infer the same data file format for all splits. Got {'train': ('json', {}), 'validation': (None, {})}.
However, running dataset = load_dataset('allenai/c4', streaming=True, data_files={'validation': 'en/c4-validation.00003-of-00008.json.gz'}, split='validation') works fine. What is the issue here?
### Steps to reproduce the bug
run code：
import os
os.environ['HF_ENDPOINT'] = 'https://hf-mirror.com'
from datasets import load_dataset
en = load_dataset("allenai/c4", "en", streaming=True)
### Expected behavior
Successfully loaded the dataset.
### Environment info
- `datasets` version: 2.18.0
- Platform: Linux-6.5.0-28-generic-x86_64-with-glibc2.17
- Python version: 3.8.19
- `huggingface_hub` version: 0.22.2
- PyArrow version: 15.0.2
- Pandas version: 2.0.3
- `fsspec` version: 2024.2.0
| false
|
2,322,980,077
|
https://api.github.com/repos/huggingface/datasets/issues/6929
|
https://github.com/huggingface/datasets/issues/6929
| 6,929
|
Avoid downloading the whole dataset when only README.me has been touched on hub.
|
open
| 2
| 2024-05-29T10:36:06
| 2024-05-29T20:51:56
| null |
zinc75
|
[
"enhancement"
] |
### Feature request
`datasets.load_dataset()` triggers a new download of the **whole dataset** when the README.md file has been touched on huggingface hub, even if data files / parquet files are the exact same.
I think the current behaviour of the load_dataset function is triggered whenever a change of the hash of latest commit on huggingface hub, but is there a clever way to only download again the dataset **if and only if** data is modified ?
### Motivation
The current behaviour is a waste of network bandwidth / disk space / research time.
### Your contribution
I don't have time to submit a PR, but I hope a simple solution will emerge from this issue !
| false
|
2,322,267,727
|
https://api.github.com/repos/huggingface/datasets/issues/6928
|
https://github.com/huggingface/datasets/pull/6928
| 6,928
|
Update process.mdx: Code Listings Fixes
|
closed
| 1
| 2024-05-29T03:17:07
| 2024-06-04T13:08:19
| 2024-06-04T12:55:00
|
FadyMorris
|
[] | null | true
|
2,322,260,725
|
https://api.github.com/repos/huggingface/datasets/issues/6927
|
https://github.com/huggingface/datasets/pull/6927
| 6,927
|
Update process.mdx: Minor Code Listings Updates and Fixes
|
closed
| 0
| 2024-05-29T03:09:01
| 2024-05-29T03:12:46
| 2024-05-29T03:12:46
|
FadyMorris
|
[] | null | true
|
2,322,164,287
|
https://api.github.com/repos/huggingface/datasets/issues/6926
|
https://github.com/huggingface/datasets/pull/6926
| 6,926
|
Update process.mdx: Fix code listing in Shard section
|
closed
| 0
| 2024-05-29T01:25:55
| 2024-05-29T03:11:20
| 2024-05-29T03:11:08
|
FadyMorris
|
[] | null | true
|
2,321,084,967
|
https://api.github.com/repos/huggingface/datasets/issues/6925
|
https://github.com/huggingface/datasets/pull/6925
| 6,925
|
Fix NonMatchingSplitsSizesError/ExpectedMoreSplits when passing data_dir/data_files in no-code Hub datasets
|
closed
| 6
| 2024-05-28T13:33:38
| 2024-11-07T20:41:58
| 2024-05-31T17:10:37
|
albertvillanova
|
[] |
Fix `NonMatchingSplitsSizesError` or `ExpectedMoreSplits` error for no-code Hub datasets if the user passes:
- `data_dir`
- `data_files`
The proposed solution is to avoid using exported dataset info (from Parquet exports) in these cases.
Additionally, also if the user passes `revision` other than "main" (so that no network requests are made).
This PR fixes a bug introduced by:
- #6714
Fix #6918, fix #6939.
| true
|
2,320,531,015
|
https://api.github.com/repos/huggingface/datasets/issues/6924
|
https://github.com/huggingface/datasets/issues/6924
| 6,924
|
Caching map result of DatasetDict.
|
open
| 3
| 2024-05-28T09:07:41
| 2025-07-28T12:57:34
| null |
MostHumble
|
[] |
Hi!
I'm currenty using the map function to tokenize a somewhat large dataset, so I need to use the cache to save ~25 mins.
Changing num_proc incduces the recomputation of the map, I'm not sure why and if this is excepted behavior?
here it says, that cached files are loaded sequentially:
https://github.com/huggingface/datasets/blob/bb2664cf540d5ce4b066365e7c8b26e7f1ca4743/src/datasets/arrow_dataset.py#L3005-L3006
it seems like I can pass in a fingerprint, and load it directly:
https://github.com/huggingface/datasets/blob/bb2664cf540d5ce4b066365e7c8b26e7f1ca4743/src/datasets/arrow_dataset.py#L3108-L3125
**Environment Setup:**
- Python 3.11.9
- datasets 2.19.1 conda-forge
- Linux 6.1.83-1.el9.elrepo.x86_64
**MRE**
```python
fixed raw_datasets
fixed tokenize_function
tokenized_datasets = raw_datasets.map(
tokenize_function,
batched=True,
num_proc=9,
remove_columns=['text'],
load_from_cache_file= True,
desc="Running tokenizer on dataset line_by_line",
)
tokenized_datasets = raw_datasets.map(
tokenize_function,
batched=True,
num_proc=5,
remove_columns=['text'],
load_from_cache_file= True,
desc="Running tokenizer on dataset line_by_line",
)
```
| false
|
2,319,292,872
|
https://api.github.com/repos/huggingface/datasets/issues/6923
|
https://github.com/huggingface/datasets/issues/6923
| 6,923
|
Export Parquet Tablet Audio-Set is null bytes in Arrow
|
open
| 0
| 2024-05-27T14:27:57
| 2024-05-27T14:27:57
| null |
anioji
|
[] |
### Describe the bug
Exporting the processed audio inside the table with the dataset.to_parquet function, the object pyarrow {bytes: null, path: "Some/Path"}
At the same time, the same dataset uploaded to the hub has bit arrays
### Steps to reproduce the bug
1.Get dataset from audio and cast it
2.Export and push dataset
3.It’s scary to be indignant at the difference in the uploaded dataset and the fact that it was saved locally
```py
from datasets import Dataset, Audio
df = Dataset.from_csv("./datasets.csv")
df = df.cast_column("audio", Audio(16000))
df.to_parquet("./datasets.parquet")
df.push_to_hub(repo_id="************", token="**********************")
```
You can use "try replicate case" for this
[replicate_packet.zip](https://github.com/huggingface/datasets/files/15457114/replicate_packet.zip)
### Expected behavior
Two parquet tables identical in content. It is obvious?
### Environment info
Python 3.11+ (I try did it in 3.12 and got same result )
| false
|
2,318,602,059
|
https://api.github.com/repos/huggingface/datasets/issues/6922
|
https://github.com/huggingface/datasets/pull/6922
| 6,922
|
Remove torchaudio remnants from code
|
closed
| 2
| 2024-05-27T08:45:07
| 2024-05-27T09:08:19
| 2024-05-27T08:59:21
|
albertvillanova
|
[] |
Remove torchaudio remnants from code.
Follow-up on:
- #5573
| true
|
2,318,394,398
|
https://api.github.com/repos/huggingface/datasets/issues/6921
|
https://github.com/huggingface/datasets/pull/6921
| 6,921
|
Support fsspec 2024.5.0
|
closed
| 2
| 2024-05-27T07:00:59
| 2024-05-27T08:07:16
| 2024-05-27T08:01:08
|
albertvillanova
|
[] |
Support fsspec 2024.5.0.
| true
|
2,317,648,021
|
https://api.github.com/repos/huggingface/datasets/issues/6920
|
https://github.com/huggingface/datasets/pull/6920
| 6,920
|
[WebDataset] Add `.pth` support for torch tensors
|
closed
| 2
| 2024-05-26T11:12:07
| 2024-05-27T09:11:17
| 2024-05-27T09:04:54
|
lhoestq
|
[] |
In this PR I add support for `.pth` but with `weights_only=True` to disallow the use of pickle
| true
|
2,315,618,993
|
https://api.github.com/repos/huggingface/datasets/issues/6919
|
https://github.com/huggingface/datasets/issues/6919
| 6,919
|
Invalid YAML in README.md: unknown tag !<tag:yaml.org,2002:python/tuple>
|
open
| 0
| 2024-05-24T14:59:45
| 2024-05-24T14:59:45
| null |
juanqui
|
[] |
### Describe the bug
I wrote a notebook to load an existing dataset, process it, and upload as a private dataset using `dataset.push_to_hub(...)` at the end. The push to hub is failing with:
```
ValueError: Invalid metadata in README.md.
- Invalid YAML in README.md: unknown tag !<tag:yaml.org,2002:python[/tuple](http://192.168.1.128:8888/tuple)> (50:11)
47 | - 4
48 | - 4
49 | - 8
50 | - !!binary |
----------------^
51 | TwAAAA==
52 | '1': !!python[/object/apply](http://192.168.1.128:8888/object/apply):nump ...
```
My dataset has a `train` and `validation` dataset. These are the features:
```
{'c1': Value(dtype='string', id=None),
'c2': Value(dtype='string', id=None),
'c3': [{'value': Value(dtype='string', id=None),
'start': Value(dtype='int64', id=None),
'end': Value(dtype='int64', id=None),
'label': Value(dtype='string', id=None)}],
'c4': Value(dtype='string', id=None),
'c5': Value(dtype='string', id=None),
'c6': Value(dtype='string', id=None),
'c7': Value(dtype='string', id=None),
'c8': Sequence(feature=Value(dtype='int32', id=None), length=-1, id=None),
'c9': Sequence(feature=Value(dtype='int8', id=None), length=-1, id=None),
'c10': Sequence(feature=Value(dtype='int8', id=None), length=-1, id=None),
'labels': Sequence(feature=ClassLabel(names=['O', 'B-ABC', 'I-ABC', ...], id=None), length=-1, id=None),
'c12': Sequence(feature=Value(dtype='string', id=None), length=-1, id=None)}
```
This used to work until I decided to cast the `labels` feature to a `Sequence(ClassLabel(...))` type with:
```
ds['train'] = ds['train'].cast_column("labels", Sequence(ClassLabel(names=list(labels))))
ds['validation'] = ds['validation'].cast_column("labels", Sequence(ClassLabel(names=list(labels))))
```
### Steps to reproduce the bug
1. Start with any token classification dataset.
2. Add a `labels` column with data such as `[0,0,0,12,13,13,13,0,0]`.
3. Cast the label column from `Sequence` to `Sequence(ClassLabel))` with:
```
labels = ['O', 'B-TEST', 'I-TEST']
ds = ds.cast_column("labels", Sequence(ClassLabel(names=labels)))
```
4. Push to hub with `ds.push_to_hub("me/awesome-stuff-dataset")`
### Expected behavior
I expected `push_to_hub` to successfully push my dataset to the hub without error.
### Environment info
Python 3.11.9
datasets==2.19.1
transformers==4.41.1
PyYAML==6.0.1
| false
|
2,315,322,738
|
https://api.github.com/repos/huggingface/datasets/issues/6918
|
https://github.com/huggingface/datasets/issues/6918
| 6,918
|
NonMatchingSplitsSizesError when using data_dir
|
closed
| 2
| 2024-05-24T12:43:39
| 2024-05-31T17:10:38
| 2024-05-31T17:10:38
|
srehaag
|
[
"bug"
] |
### Describe the bug
Loading a dataset from with a data_dir argument generates a NonMatchingSplitsSizesError if there are multiple directories in the dataset.
This appears to happen because the expected split is calculated based on the data in all the directories whereas the recorded split is calculated based on the data in the directory specified using the data_dir argument.
This is recent behavior. Until the past few weeks loading using the data_dir argument worked without any issue.
### Steps to reproduce the bug
Simple test dataset available here: https://huggingface.co/datasets/srehaag/hf-bug-temp
The dataset contains two directories "data1" and "data2", each with a file called "train.parquet" with a 2 x 5 table.
from datasets import load_dataset
dataset = load_dataset("srehaag/hf-bug-temp", data_dir = "data1")
Generates:
---------------------------------------------------------------------------
NonMatchingSplitsSizesError Traceback (most recent call last)
Cell In[3], <a href='vscode-notebook-cell:?execution_count=3&line=2'>line 2</a>
<a href='vscode-notebook-cell:?execution_count=3&line=1'>1</a> from datasets import load_dataset
----> <a href='vscode-notebook-cell:?execution_count=3&line=2'>2</a> dataset = load_dataset("srehaag/hf-bug-temp", data_dir = "data1")
File ~/.python/current/lib/python3.10/site-packages/datasets/load.py:2609, in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, verification_mode, ignore_verifications, keep_in_memory, save_infos, revision, token, use_auth_token, task, streaming, num_proc, storage_options, trust_remote_code, **config_kwargs)
<a href='~/.python/current/lib/python3.10/site-packages/datasets/load.py:2606'>2606</a> return builder_instance.as_streaming_dataset(split=split)
<a href='~/.python/current/lib/python3.10/site-packages/datasets/load.py:2608'>2608</a> # Download and prepare data
-> <a href='~/.python/current/lib/python3.10/site-packages/datasets/load.py:2609'>2609</a> builder_instance.download_and_prepare(
<a href='~/.python/current/lib/python3.10/site-packages/datasets/load.py:2610'>2610</a> download_config=download_config,
<a href='~/.python/current/lib/python3.10/site-packages/datasets/load.py:2611'>2611</a> download_mode=download_mode,
<a href='~/.python/current/lib/python3.10/site-packages/datasets/load.py:2612'>2612</a> verification_mode=verification_mode,
<a href='~/.python/current/lib/python3.10/site-packages/datasets/load.py:2613'>2613</a> num_proc=num_proc,
<a href='~/.python/current/lib/python3.10/site-packages/datasets/load.py:2614'>2614</a> storage_options=storage_options,
<a href='~/.python/current/lib/python3.10/site-packages/datasets/load.py:2615'>2615</a> )
<a href='~/.python/current/lib/python3.10/site-packages/datasets/load.py:2617'>2617</a> # Build dataset for splits
<a href='~/.python/current/lib/python3.10/site-packages/datasets/load.py:2618'>2618</a> keep_in_memory = (
<a href='~/.python/current/lib/python3.10/site-packages/datasets/load.py:2619'>2619</a> keep_in_memory if keep_in_memory is not None else is_small_dataset(builder_instance.info.dataset_size)
<a href='~/.python/current/lib/python3.10/site-packages/datasets/load.py:2620'>2620</a> )
File ~/.python/current/lib/python3.10/site-packages/datasets/builder.py:1027, in DatasetBuilder.download_and_prepare(self, output_dir, download_config, download_mode, verification_mode, ignore_verifications, try_from_hf_gcs, dl_manager, base_path, use_auth_token, file_format, max_shard_size, num_proc, storage_options, **download_and_prepare_kwargs)
<a href='~/.python/current/lib/python3.10/site-packages/datasets/builder.py:1025'>1025</a> if num_proc is not None:
<a href='~/.python/current/lib/python3.10/site-packages/datasets/builder.py:1026'>1026</a> prepare_split_kwargs["num_proc"] = num_proc
-> <a href='~/.python/current/lib/python3.10/site-packages/datasets/builder.py:1027'>1027</a> self._download_and_prepare(
<a href='~/.python/current/lib/python3.10/site-packages/datasets/builder.py:1028'>1028</a> dl_manager=dl_manager,
<a href='~/.python/current/lib/python3.10/site-packages/datasets/builder.py:1029'>1029</a> verification_mode=verification_mode,
<a href='~/.python/current/lib/python3.10/site-packages/datasets/builder.py:1030'>1030</a> **prepare_split_kwargs,
<a href='~/.python/current/lib/python3.10/site-packages/datasets/builder.py:1031'>1031</a> **download_and_prepare_kwargs,
<a href='~/.python/current/lib/python3.10/site-packages/datasets/builder.py:1032'>1032</a> )
<a href='~/.python/current/lib/python3.10/site-packages/datasets/builder.py:1033'>1033</a> # Sync info
<a href='~/.python/current/lib/python3.10/site-packages/datasets/builder.py:1034'>1034</a> self.info.dataset_size = sum(split.num_bytes for split in self.info.splits.values())
File ~/.python/current/lib/python3.10/site-packages/datasets/builder.py:1140, in DatasetBuilder._download_and_prepare(self, dl_manager, verification_mode, **prepare_split_kwargs)
<a href='~/.python/current/lib/python3.10/site-packages/datasets/builder.py:1137'>1137</a> dl_manager.manage_extracted_files()
<a href='~/.python/current/lib/python3.10/site-packages/datasets/builder.py:1139'>1139</a> if verification_mode == VerificationMode.BASIC_CHECKS or verification_mode == VerificationMode.ALL_CHECKS:
-> <a href='~/.python/current/lib/python3.10/site-packages/datasets/builder.py:1140'>1140</a> verify_splits(self.info.splits, split_dict)
<a href='~/.python/current/lib/python3.10/site-packages/datasets/builder.py:1142'>1142</a> # Update the info object with the splits.
<a href='~/.python/current/lib/python3.10/site-packages/datasets/builder.py:1143'>1143</a> self.info.splits = split_dict
File ~/.python/current/lib/python3.10/site-packages/datasets/utils/info_utils.py:101, in verify_splits(expected_splits, recorded_splits)
<a href='~/.python/current/lib/python3.10/site-packages/datasets/utils/info_utils.py:95'>95</a> bad_splits = [
<a href='~/.python/current/lib/python3.10/site-packages/datasets/utils/info_utils.py:96'>96</a> {"expected": expected_splits[name], "recorded": recorded_splits[name]}
<a href='~/.python/current/lib/python3.10/site-packages/datasets/utils/info_utils.py:97'>97</a> for name in expected_splits
<a href='~/.python/current/lib/python3.10/site-packages/datasets/utils/info_utils.py:98'>98</a> if expected_splits[name].num_examples != recorded_splits[name].num_examples
<a href='~/.python/current/lib/python3.10/site-packages/datasets/utils/info_utils.py:99'>99</a> ]
<a href='~/.python/current/lib/python3.10/site-packages/datasets/utils/info_utils.py:100'>100</a> if len(bad_splits) > 0:
--> <a href='~/.python/current/lib/python3.10/site-packages/datasets/utils/info_utils.py:101'>101</a> raise NonMatchingSplitsSizesError(str(bad_splits))
<a href='~/.python/current/lib/python3.10/site-packages/datasets/utils/info_utils.py:102'>102</a> logger.info("All the splits matched successfully.")
NonMatchingSplitsSizesError: [{'expected': SplitInfo(name='train', num_bytes=212, num_examples=10, shard_lengths=None, dataset_name=None), 'recorded': SplitInfo(name='train', num_bytes=106, num_examples=5, shard_lengths=None, dataset_name='hf-bug-temp')}]
__________
By contrast, this loads the data from both data1/train.parquet and data2/train.parquet without any error message:
from datasets import load_dataset
dataset = load_dataset("srehaag/hf-bug-temp")
### Expected behavior
Should load the 5 x 2 table from data1/train.parquet without error message.
### Environment info
Used Codespaces to simplify environment (see details below), but bug is present across various configurations.
- `datasets` version: 2.19.1
- Platform: Linux-6.5.0-1021-azure-x86_64-with-glibc2.31
- Python version: 3.10.13
- `huggingface_hub` version: 0.23.1
- PyArrow version: 16.1.0
- Pandas version: 2.2.2
- `fsspec` version: 2024.3.1
| false
|
2,314,683,663
|
https://api.github.com/repos/huggingface/datasets/issues/6917
|
https://github.com/huggingface/datasets/issues/6917
| 6,917
|
WinError 32 The process cannot access the file during load_dataset
|
open
| 0
| 2024-05-24T07:54:51
| 2024-05-24T07:54:51
| null |
elwe-2808
|
[] |
### Describe the bug
When I try to load the opus_book from hugging face (following the [guide on the website](https://huggingface.co/docs/transformers/main/en/tasks/translation))
```python
from datasets import load_dataset, Dataset
dataset = load_dataset("Helsinki-NLP/opus_books", "en-fr", features=["id", "translation"])
```
I get an error:
`PermissionError: [WinError 32] The process cannot access the file because it is being used by another process: 'C:/Users/Me/.cache/huggingface/datasets/Helsinki-NLP___parquet/ca-de-a39f1ef185b9b73b/0.0.0/2a3b91fbd88a2c90d1dbbb32b460cf621d31bd5b05b934492fdef7d8d6f236ec.incomplete\\parquet-train-00000-00000-of-NNNNN.arrow'
`
<details><summary>Full stacktrace</summary>
<p>
```python
AttributeError Traceback (most recent call last)
File c:\Users\Me\.conda\envs\ia\lib\site-packages\datasets\builder.py:1858, in ArrowBasedBuilder._prepare_split_single(self, gen_kwargs, fpath, file_format, max_shard_size, job_id)
[1857](file:///C:/Users/Me/.conda/envs/ia/lib/site-packages/datasets/builder.py:1857) _time = time.time()
-> [1858](file:///C:/Users/Me/.conda/envs/ia/lib/site-packages/datasets/builder.py:1858) for _, table in generator:
[1859](file:///C:/Users/Me/.conda/envs/ia/lib/site-packages/datasets/builder.py:1859) if max_shard_size is not None and writer._num_bytes > max_shard_size:
File c:\Users\Me\.conda\envs\ia\lib\site-packages\datasets\packaged_modules\parquet\parquet.py:59, in Parquet._generate_tables(self, files)
[58](file:///C:/Users/Me/.conda/envs/ia/lib/site-packages/datasets/packaged_modules/parquet/parquet.py:58) def _generate_tables(self, files):
---> [59](file:///C:/Users/Me/.conda/envs/ia/lib/site-packages/datasets/packaged_modules/parquet/parquet.py:59) schema = self.config.features.arrow_schema if self.config.features is not None else None
[60](file:///C:/Users/Me/.conda/envs/ia/lib/site-packages/datasets/packaged_modules/parquet/parquet.py:60) if self.config.features is not None and self.config.columns is not None:
AttributeError: 'list' object has no attribute 'arrow_schema'
During handling of the above exception, another exception occurred:
AttributeError Traceback (most recent call last)
File c:\Users\Me\.conda\envs\ia\lib\site-packages\datasets\builder.py:1882, in ArrowBasedBuilder._prepare_split_single(self, gen_kwargs, fpath, file_format, max_shard_size, job_id)
[1881](file:///C:/Users/Me/.conda/envs/ia/lib/site-packages/datasets/builder.py:1881) num_shards = shard_id + 1
-> [1882](file:///C:/Users/Me/.conda/envs/ia/lib/site-packages/datasets/builder.py:1882) num_examples, num_bytes = writer.finalize()
[1883](file:///C:/Users/Me/.conda/envs/ia/lib/site-packages/datasets/builder.py:1883) writer.close()
File c:\Users\Me\.conda\envs\ia\lib\site-packages\datasets\arrow_writer.py:584, in ArrowWriter.finalize(self, close_stream)
[583](file:///C:/Users/Me/.conda/envs/ia/lib/site-packages/datasets/arrow_writer.py:583) # If schema is known, infer features even if no examples were written
--> [584](file:///C:/Users/Me/.conda/envs/ia/lib/site-packages/datasets/arrow_writer.py:584) if self.pa_writer is None and self.schema:
...
--> [627](file:///C:/Users/Me/.conda/envs/ia/lib/shutil.py:627) os.unlink(fullname)
[628](file:///C:/Users/Me/.conda/envs/ia/lib/shutil.py:628) except OSError:
[629](file:///C:/Users/Me/.conda/envs/ia/lib/shutil.py:629) onerror(os.unlink, fullname, sys.exc_info())
PermissionError: [WinError 32] The process cannot access the file because it is being used by another process: 'C:/Users/Me/.cache/huggingface/datasets/Helsinki-NLP___parquet/ca-de-a39f1ef185b9b73b/0.0.0/2a3b91fbd88a2c90d1dbbb32b460cf621d31bd5b05b934492fdef7d8d6f236ec.incomplete\\parquet-train-00000-00000-of-NNNNN.arrow'
```
</p>
</details>
### Steps to reproduce the bug
Steps to reproduce:
Just execute these lines
```python
from datasets import load_dataset, Dataset
dataset = load_dataset("Helsinki-NLP/opus_books", "en-fr", features=["id", "translation"])
```
### Expected behavior
I expect the dataset to be loaded without any errors.
### Environment info
| Package| Version|
|--------|--------|
| transformers| 4.37.2|
| python| 3.9.19|
| pytorch| 2.3.0|
| datasets|2.12.0 |
| arrow | 1.2.3|
I am using Conda on Windows 11.
| false
|
2,311,675,564
|
https://api.github.com/repos/huggingface/datasets/issues/6916
|
https://github.com/huggingface/datasets/issues/6916
| 6,916
|
```push_to_hub()``` - Prevent Automatic Generation of Splits
|
closed
| 0
| 2024-05-22T23:52:15
| 2024-05-23T00:07:53
| 2024-05-23T00:07:53
|
jetlime
|
[] |
### Describe the bug
I currently have a dataset which has not been splited. When pushing the dataset to my hugging face dataset repository, it is split into a testing and training set. How can I prevent the split from happening?
### Steps to reproduce the bug
1. Have a unsplit dataset
```python
Dataset({ features: ['input', 'output', 'Attack', '__index_level_0__'], num_rows: 944685 })
```
2. Push it to huggingface
```python
dataset.push_to_hub(dataset_name)
```
3. On the hugging face dataset repo, the dataset then appears to be splited:
4. Indeed, when loading the dataset from this repo, the dataset is split in two testing and training set.
```python
from datasets import load_dataset, Dataset
dataset = load_dataset("Jetlime/NF-CSE-CIC-IDS2018-v2", streaming=True)
dataset
```
output:
```
IterableDatasetDict({
train: IterableDataset({
features: ['input', 'output', 'Attack', '__index_level_0__'],
n_shards: 2
})
test: IterableDataset({
features: ['input', 'output', 'Attack', '__index_level_0__'],
n_shards: 1
})
```
### Expected behavior
The dataset shall not be splited, as not requested.
### Environment info
- `datasets` version: 2.19.1
- Platform: Linux-6.2.0-35-generic-x86_64-with-glibc2.35
- Python version: 3.10.12
- `huggingface_hub` version: 0.23.0
- PyArrow version: 15.0.2
- Pandas version: 2.2.2
- `fsspec` version: 2024.3.1
| false
|
2,310,564,961
|
https://api.github.com/repos/huggingface/datasets/issues/6915
|
https://github.com/huggingface/datasets/pull/6915
| 6,915
|
Validate config name and data_files in packaged modules
|
closed
| 5
| 2024-05-22T13:36:33
| 2024-06-06T09:32:10
| 2024-06-06T09:24:35
|
albertvillanova
|
[] |
Validate the config attributes `name` and `data_files` in packaged modules by making the derived classes call their parent `__post_init__` method.
Note that their parent `BuilderConfig` validates its attributes `name` and `data_files` in its `__post_init__` method: https://github.com/huggingface/datasets/blob/60d21efbc01e15d0b596ac1072750cbecd91548a/src/datasets/builder.py#L128-L137
This PR makes the derived config classes call their parent `__post_init__` method to validate their `name` and `data_files` attributes.
| true
|
2,310,107,326
|
https://api.github.com/repos/huggingface/datasets/issues/6914
|
https://github.com/huggingface/datasets/pull/6914
| 6,914
|
Preserve JSON column order and support list of strings field
|
closed
| 2
| 2024-05-22T09:58:54
| 2024-05-29T13:18:47
| 2024-05-29T13:12:23
|
albertvillanova
|
[] |
Preserve column order when loading from a JSON file with a list of dict (or with a field containing a list of dicts).
Additionally, support JSON file with a list of strings field.
Fix #6913.
| true
|
2,309,605,889
|
https://api.github.com/repos/huggingface/datasets/issues/6913
|
https://github.com/huggingface/datasets/issues/6913
| 6,913
|
Column order is nondeterministic when loading from JSON
|
closed
| 0
| 2024-05-22T05:30:14
| 2024-05-29T13:12:24
| 2024-05-29T13:12:24
|
albertvillanova
|
[
"bug"
] |
As reported by @meg-huggingface, the order of the JSON object keys is not preserved while loading a dataset from a JSON file with a list of objects.
For example, when loading a JSON files with a list of objects, each with the following ordered keys:
- [ID, Language, Topic],
the resulting dataset may have columns:
- [ID, Topic, Language], or
- [Topic, Language, ID], or
- [Topic, ID, Language],...
This issue is caused by the use of a Python set (which does not preserve the order):
https://github.com/huggingface/datasets/blob/60d21efbc01e15d0b596ac1072750cbecd91548a/src/datasets/packaged_modules/json/json.py#L168
introduced in
- #5772
| false
|
2,309,365,961
|
https://api.github.com/repos/huggingface/datasets/issues/6912
|
https://github.com/huggingface/datasets/issues/6912
| 6,912
|
Add MedImg for streaming
|
open
| 8
| 2024-05-22T00:55:30
| 2024-09-05T16:53:54
| null |
lhallee
|
[
"dataset request"
] |
### Feature request
Host the MedImg dataset (similar to Imagenet but for biomedical images).
### Motivation
There is a clear need for biomedical image foundation models and large scale biomedical datasets that are easily streamable. This would be an excellent tool for the biomedical community.
### Your contribution
MedImg can be found [here](https://www.cuilab.cn/medimg/#).
| false
|
2,308,152,711
|
https://api.github.com/repos/huggingface/datasets/issues/6911
|
https://github.com/huggingface/datasets/pull/6911
| 6,911
|
Remove dead code for non-dict data_files from packaged modules
|
closed
| 2
| 2024-05-21T12:10:24
| 2024-05-23T08:05:58
| 2024-05-23T07:59:57
|
albertvillanova
|
[] |
Remove dead code for non-dict data_files from packaged modules.
Since the merge of this PR:
- #2986
the builders' variable self.config.data_files is always a dict, which makes the condition on (str, list, tuple) dead code.
| true
|
2,307,570,084
|
https://api.github.com/repos/huggingface/datasets/issues/6910
|
https://github.com/huggingface/datasets/pull/6910
| 6,910
|
Fix wrong type hints in data_files
|
closed
| 2
| 2024-05-21T07:41:09
| 2024-05-23T06:04:05
| 2024-05-23T05:58:05
|
albertvillanova
|
[] |
Fix wrong type hints in data_files introduced in:
- #6493
| true
|
2,307,508,120
|
https://api.github.com/repos/huggingface/datasets/issues/6909
|
https://github.com/huggingface/datasets/pull/6909
| 6,909
|
Update requests >=2.32.1 to fix vulnerability
|
closed
| 2
| 2024-05-21T07:11:20
| 2024-05-21T07:45:58
| 2024-05-21T07:38:25
|
albertvillanova
|
[] |
Update requests >=2.32.1 to fix vulnerability.
| true
|
2,304,958,116
|
https://api.github.com/repos/huggingface/datasets/issues/6908
|
https://github.com/huggingface/datasets/issues/6908
| 6,908
|
Fail to load "stas/c4-en-10k" dataset since 2.16 version
|
closed
| 2
| 2024-05-20T02:43:59
| 2024-05-24T10:58:09
| 2024-05-24T10:58:09
|
guch8017
|
[] |
### Describe the bug
When update datasets library to version 2.16+ ( I test it on 2.16, 2.19.0 and 2.19.1), using the following code to load stas/c4-en-10k dataset
```python
from datasets import load_dataset, Dataset
dataset = load_dataset('stas/c4-en-10k')
```
and then it raise UnicodeDecodeError like
```
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/*/conda3/envs/watermark/lib/python3.10/site-packages/datasets/load.py", line 2523, in load_dataset
builder_instance = load_dataset_builder(
File "/home/*/conda3/envs/watermark/lib/python3.10/site-packages/datasets/load.py", line 2195, in load_dataset_builder
dataset_module = dataset_module_factory(
File "/home/*/conda3/envs/watermark/lib/python3.10/site-packages/datasets/load.py", line 1846, in dataset_module_factory
raise e1 from None
File "/home/*/conda3/envs/watermark/lib/python3.10/site-packages/datasets/load.py", line 1798, in dataset_module_factory
can_load_config_from_parquet_export = "DEFAULT_CONFIG_NAME" not in f.read()
File "/home/*/conda3/envs/watermark/lib/python3.10/codecs.py", line 322, in decode
(result, consumed) = self._buffer_decode(data, self.errors, final)
UnicodeDecodeError: 'utf-8' codec can't decode byte 0x8b in position 1: invalid start byte
```
I found that fs.open loads a gzip file and parses it like plain text using utf-8 encoder.
```python
fs = HfFileSystem('https://huggingface.co')
fs.open("datasets/stas/c4-en-10k/c4-en-10k.py", "rb")
data = fs.read() # data is gzip bytes begin with b'\x1f\x8b\x08\x00\x00\tn\x88\x00...'
data2 = unzip_gzip_bytes(data) # data2 is what we want: '# coding=utf-8\n# Copyright 2020 The HuggingFace Datasets...'
```
### Steps to reproduce the bug
1. Install datasets between version 2.16 and 2.19
2. Use `datasets.load_dataset` method to load `stas/c4-en-10k` dataset.
### Expected behavior
Load dataset normally.
### Environment info
Platform = Linux-5.4.0-159-generic-x86_64-with-glibc2.35
Python = 3.10.14
Datasets = 2.19
| false
|
2,303,855,833
|
https://api.github.com/repos/huggingface/datasets/issues/6907
|
https://github.com/huggingface/datasets/issues/6907
| 6,907
|
Support the deserialization of json lines files comprised of lists
|
open
| 1
| 2024-05-18T05:07:23
| 2024-05-18T08:53:28
| null |
umarbutler
|
[
"enhancement"
] |
### Feature request
I manage a somewhat large and popular Hugging Face dataset known as the [Open Australian Legal Corpus](https://huggingface.co/datasets/umarbutler/open-australian-legal-corpus). I recently updated my corpus to be stored in a json lines file where each line is an array and each element represents a value at a particular column. Previously, my corpus was stored as a json lines file where each line was a dictionary and the keys were the fields.
Essentially, a line in my json lines file used to look like this:
```json
{"version_id":"","type":"","jurisdiction":"","source":"","citation":"","url":"","when_scraped":"","text":""}
```
And now it looks like this:
```json
["","","","","","","",""]
```
This saves 65 bytes per document and allows me very quickly serialise and deserialise documents via `msgspec`.
After making this change, I found that `datasets` was incapable of deserialising my Corpus without a custom loading script, even if I ensured that the `dataset_info` field in my dataset card contained the desired names of my features.
I would like to request that functionality be added to support this format which is more memory-efficent and faster than using dictionaries.
### Motivation
The [documentation](https://huggingface.co/docs/datasets/en/dataset_script) for creating dataset loading scripts asserts that:
> In the next major release, the new safety features of 🤗 Datasets will disable running dataset loading scripts by default, and you will have to pass trust_remote_code=True to load datasets that require running a dataset script.
I would rather not require my users to pass `trust_remote_code=True` which means that I will need built-in support for this format.
### Your contribution
I would be happy to submit a PR for this if this is something you would incorporate into `datasets` and if I can be pointed to where the code would need to go.
| false
|
2,303,679,119
|
https://api.github.com/repos/huggingface/datasets/issues/6906
|
https://github.com/huggingface/datasets/issues/6906
| 6,906
|
irc_disentangle - Issue with splitting data
|
closed
| 6
| 2024-05-17T23:19:37
| 2024-07-16T00:21:56
| 2024-07-08T06:18:08
|
eor51355
|
[] |
### Describe the bug
I am trying to access your database through python using "datasets.load_dataset("irc_disentangle")" and I am getting this error message:
ValueError: Instruction "train" corresponds to no data!
### Steps to reproduce the bug
import datasets
ds = datasets.load_dataset('irc_disentangle')
ds
### Expected behavior
The data is supposed to load into ds and be accessable as such:
ds['train'][1050], ds['train'][1055]
### Environment info
I tired Python 3.12 and 3.10
| false
|
2,303,098,587
|
https://api.github.com/repos/huggingface/datasets/issues/6905
|
https://github.com/huggingface/datasets/issues/6905
| 6,905
|
Extraction protocol for arrow files is not defined
|
closed
| 1
| 2024-05-17T16:01:41
| 2025-02-06T19:50:22
| 2025-02-06T19:50:20
|
radulescupetru
|
[] |
### Describe the bug
Passing files with `.arrow` extension into data_files argument, at least when `streaming=True` is very slow.
### Steps to reproduce the bug
Basically it goes through the `_get_extraction_protocol` method located [here](https://github.com/huggingface/datasets/blob/main/src/datasets/utils/file_utils.py#L820)
The method then looks at some base known extensions where `arrow` is not defined so it proceeds to determine the compression with the magic number method which is slow when dealing with a lot of files which are stored in s3 and by looking at this predefined list, I don't see `arrow` in there either so in the end it return None:
```
MAGIC_NUMBER_TO_COMPRESSION_PROTOCOL = {
bytes.fromhex("504B0304"): "zip",
bytes.fromhex("504B0506"): "zip", # empty archive
bytes.fromhex("504B0708"): "zip", # spanned archive
bytes.fromhex("425A68"): "bz2",
bytes.fromhex("1F8B"): "gzip",
bytes.fromhex("FD377A585A00"): "xz",
bytes.fromhex("04224D18"): "lz4",
bytes.fromhex("28B52FFD"): "zstd",
}
```
### Expected behavior
My expectation is that `arrow` would be in the known lists so it would return None without going through the magic number method.
### Environment info
datasets 2.19.0
| false
|
2,302,912,179
|
https://api.github.com/repos/huggingface/datasets/issues/6904
|
https://github.com/huggingface/datasets/pull/6904
| 6,904
|
Fix decoding multi part extension
|
closed
| 3
| 2024-05-17T14:32:57
| 2024-05-17T14:52:56
| 2024-05-17T14:46:54
|
lhoestq
|
[] |
e.g. a field named `url.txt` should be a treated as text
I also included a small fix to support .npz correctly
| true
|
2,300,436,053
|
https://api.github.com/repos/huggingface/datasets/issues/6903
|
https://github.com/huggingface/datasets/issues/6903
| 6,903
|
Add the option of saving in parquet instead of arrow
|
open
| 18
| 2024-05-16T13:35:51
| 2025-05-19T12:14:14
| null |
arita37
|
[
"enhancement"
] |
### Feature request
In dataset.save_to_disk('/path/to/save/dataset'),
add the option to save in parquet format
dataset.save_to_disk('/path/to/save/dataset', format="parquet"),
because arrow is not used for Production Big data.... (only parquet)
### Motivation
because arrow is not used for Production Big data.... (only parquet)
### Your contribution
I can do the testing !
| false
|
2,300,256,241
|
https://api.github.com/repos/huggingface/datasets/issues/6902
|
https://github.com/huggingface/datasets/pull/6902
| 6,902
|
Make CLI convert_to_parquet not raise error if no rights to create script branch
|
closed
| 2
| 2024-05-16T12:21:27
| 2024-06-03T04:43:17
| 2024-05-16T12:51:05
|
albertvillanova
|
[] |
Make CLI convert_to_parquet not raise error if no rights to create "script" branch.
Not that before this PR, the error was not critical because it was raised at the end of the script, once all the rest of the steps were already performed.
Fix #6901.
Bug introduced in datasets-2.19.0 by:
- #6809
| true
|
2,300,167,465
|
https://api.github.com/repos/huggingface/datasets/issues/6901
|
https://github.com/huggingface/datasets/issues/6901
| 6,901
|
HTTPError 403 raised by CLI convert_to_parquet when creating script branch on 3rd party repos
|
closed
| 0
| 2024-05-16T11:40:22
| 2024-05-16T12:51:06
| 2024-05-16T12:51:06
|
albertvillanova
|
[
"bug"
] |
CLI convert_to_parquet cannot create "script" branch on 3rd party repos.
It can only create it on repos where the user executing the script has write access.
Otherwise, a 403 Forbidden HTTPError is raised:
```
Traceback (most recent call last):
File "/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/_errors.py", line 304, in hf_raise_for_status
response.raise_for_status()
File "/usr/local/lib/python3.10/dist-packages/requests/models.py", line 1021, in raise_for_status
raise HTTPError(http_error_msg, response=self)
requests.exceptions.HTTPError: 403 Client Error: Forbidden for url: https://huggingface.co/api/datasets/ORG/DATASET/branch/script
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/usr/local/bin/datasets-cli", line 8, in <module>
sys.exit(main())
File "/usr/local/lib/python3.10/dist-packages/datasets/commands/datasets_cli.py", line 41, in main
service.run()
File "/usr/local/lib/python3.10/dist-packages/datasets/commands/convert_to_parquet.py", line 92, in run
create_branch(dataset_id, branch="script", repo_type="dataset", token=token, exist_ok=True)
File "/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/_validators.py", line 114, in _inner_fn
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/huggingface_hub/hf_api.py", line 5503, in create_branch
hf_raise_for_status(response)
File "/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/_errors.py", line 367, in hf_raise_for_status
raise HfHubHTTPError(message, response=response) from e
huggingface_hub.utils._errors.HfHubHTTPError: (Request ID: Root=1-6645ee0d-4db1ed8a1fbe04956be15897;139a6e23-df7d-4f62-b5ba-adb6d8e6e696)
403 Forbidden: Forbidden: cannot write to script.
Cannot access content at: https://huggingface.co/api/datasets/ORG/DATASET/branch/script.
If you are trying to create or update content,make sure you have a token with the `write` role.
```
| false
|
2,298,489,733
|
https://api.github.com/repos/huggingface/datasets/issues/6900
|
https://github.com/huggingface/datasets/issues/6900
| 6,900
|
[WebDataset] KeyError with user-defined `Features` when a field is missing in an example
|
closed
| 5
| 2024-05-15T17:48:34
| 2024-06-28T09:30:13
| 2024-06-28T09:30:13
|
lhoestq
|
[] |
reported at https://huggingface.co/datasets/ProGamerGov/synthetic-dataset-1m-dalle3-high-quality-captions/discussions/1
```
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 109, in _generate_examples
example[field_name] = {"path": example["__key__"] + "." + field_name, "bytes": example[field_name]}
```
| false
|
2,298,059,597
|
https://api.github.com/repos/huggingface/datasets/issues/6899
|
https://github.com/huggingface/datasets/issues/6899
| 6,899
|
List of dictionary features get standardized
|
open
| 2
| 2024-05-15T14:11:35
| 2025-04-01T20:48:03
| null |
sohamparikh
|
[] |
### Describe the bug
Hi, i’m trying to create a HF dataset from a list using Dataset.from_list.
Each sample in the list is a dict with the same keys (which will be my features). The values for each feature are a list of dictionaries, and each such dictionary has a different set of keys. However, the datasets library standardizes all dictionaries under a feature and adds all possible keys (with None value) from all the dictionaries under that feature.
How can I keep the same set of keys as in the original list for each dictionary under a feature?
### Steps to reproduce the bug
```
from datasets import Dataset
# Define a function to generate a sample with "tools" feature
def generate_sample():
# Generate random sample data
sample_data = {
"text": "Sample text",
"feature_1": []
}
# Add feature_1 with random keys for this sample
feature_1 = [{"key1": "value1"}, {"key2": "value2"}] # Example feature_1 with random keys
sample_data["feature_1"].extend(feature_1)
return sample_data
# Generate multiple samples
num_samples = 10
samples = [generate_sample() for _ in range(num_samples)]
# Create a Hugging Face Dataset
dataset = Dataset.from_list(samples)
dataset[0]
```
```{'text': 'Sample text', 'feature_1': [{'key1': 'value1', 'key2': None}, {'key1': None, 'key2': 'value2'}]}```
### Expected behavior
```{'text': 'Sample text', 'feature_1': [{'key1': 'value1'}, {'key2': 'value2'}]}```
### Environment info
- `datasets` version: 2.19.1
- Platform: Linux-5.15.0-1040-nvidia-x86_64-with-glibc2.35
- Python version: 3.10.13
- `huggingface_hub` version: 0.23.0
- PyArrow version: 15.0.0
- Pandas version: 2.2.0
- `fsspec` version: 2023.10.0
| false
|
2,294,432,108
|
https://api.github.com/repos/huggingface/datasets/issues/6898
|
https://github.com/huggingface/datasets/pull/6898
| 6,898
|
Fix YAML error in README files appearing on GitHub
|
closed
| 3
| 2024-05-14T05:21:57
| 2024-05-16T14:36:57
| 2024-05-16T14:28:16
|
albertvillanova
|
[] |
Fix YAML error in README files appearing on GitHub.
See error message:
Fix #6897.
| true
|
2,293,428,243
|
https://api.github.com/repos/huggingface/datasets/issues/6897
|
https://github.com/huggingface/datasets/issues/6897
| 6,897
|
datasets template guide :: issue in documentation YAML
|
closed
| 2
| 2024-05-13T17:33:59
| 2024-05-16T14:28:17
| 2024-05-16T14:28:17
|
bghira
|
[] |
### Describe the bug
There is a YAML error at the top of the page, and I don't think it's supposed to be there
### Steps to reproduce the bug
1. Browse to [this tutorial document](https://github.com/huggingface/datasets/blob/main/templates/README_guide.md)
2. Observe a big red error at the top
3. The rest of the document remains functional
### Expected behavior
I think the YAML block should be displayed or ignored.
### Environment info
N/A
| false
|
2,293,176,061
|
https://api.github.com/repos/huggingface/datasets/issues/6896
|
https://github.com/huggingface/datasets/issues/6896
| 6,896
|
Regression bug: `NonMatchingSplitsSizesError` for (possibly) overwritten dataset
|
open
| 1
| 2024-05-13T15:41:57
| 2025-03-25T01:21:06
| null |
finiteautomata
|
[] |
### Describe the bug
While trying to load the dataset `https://huggingface.co/datasets/pysentimiento/spanish-tweets-small`, I get this error:
```python
---------------------------------------------------------------------------
NonMatchingSplitsSizesError Traceback (most recent call last)
[<ipython-input-1-d6a3c721d3b8>](https://localhost:8080/#) in <cell line: 3>()
1 from datasets import load_dataset
2
----> 3 ds = load_dataset("pysentimiento/spanish-tweets-small")
3 frames
[/usr/local/lib/python3.10/dist-packages/datasets/load.py](https://localhost:8080/#) in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, verification_mode, ignore_verifications, keep_in_memory, save_infos, revision, token, use_auth_token, task, streaming, num_proc, storage_options, **config_kwargs)
2150
2151 # Download and prepare data
-> 2152 builder_instance.download_and_prepare(
2153 download_config=download_config,
2154 download_mode=download_mode,
[/usr/local/lib/python3.10/dist-packages/datasets/builder.py](https://localhost:8080/#) in download_and_prepare(self, output_dir, download_config, download_mode, verification_mode, ignore_verifications, try_from_hf_gcs, dl_manager, base_path, use_auth_token, file_format, max_shard_size, num_proc, storage_options, **download_and_prepare_kwargs)
946 if num_proc is not None:
947 prepare_split_kwargs["num_proc"] = num_proc
--> 948 self._download_and_prepare(
949 dl_manager=dl_manager,
950 verification_mode=verification_mode,
[/usr/local/lib/python3.10/dist-packages/datasets/builder.py](https://localhost:8080/#) in _download_and_prepare(self, dl_manager, verification_mode, **prepare_split_kwargs)
1059
1060 if verification_mode == VerificationMode.BASIC_CHECKS or verification_mode == VerificationMode.ALL_CHECKS:
-> 1061 verify_splits(self.info.splits, split_dict)
1062
1063 # Update the info object with the splits.
[/usr/local/lib/python3.10/dist-packages/datasets/utils/info_utils.py](https://localhost:8080/#) in verify_splits(expected_splits, recorded_splits)
98 ]
99 if len(bad_splits) > 0:
--> 100 raise NonMatchingSplitsSizesError(str(bad_splits))
101 logger.info("All the splits matched successfully.")
102
NonMatchingSplitsSizesError: [{'expected': SplitInfo(name='train', num_bytes=82649695458, num_examples=597433111, shard_lengths=None, dataset_name=None), 'recorded': SplitInfo(name='train', num_bytes=3358310095, num_examples=24898932, shard_lengths=[3626991, 3716991, 4036990, 3506990, 3676990, 3716990, 2616990], dataset_name='spanish-tweets-small')}]
```
I think I had this dataset updated, might be related to #6271
It is working fine as late in `2.10.0` , but not in `2.13.0` onwards.
### Steps to reproduce the bug
```python
from datasets import load_dataset
ds = load_dataset("pysentimiento/spanish-tweets-small")
```
You can run it in [this notebook](https://colab.research.google.com/drive/1FdhqLiVimHIlkn7B54DbhizeQ4U3vGVl#scrollTo=YgA50cBSibUg)
### Expected behavior
Load the dataset without any error
### Environment info
- `datasets` version: 2.13.0
- Platform: Linux-6.1.58+-x86_64-with-glibc2.35
- Python version: 3.10.12
- Huggingface_hub version: 0.20.3
- PyArrow version: 14.0.2
- Pandas version: 2.0.3
| false
|
2,292,993,156
|
https://api.github.com/repos/huggingface/datasets/issues/6895
|
https://github.com/huggingface/datasets/pull/6895
| 6,895
|
Document that to_json defaults to JSON Lines
|
closed
| 2
| 2024-05-13T14:22:34
| 2024-05-16T14:37:25
| 2024-05-16T14:31:26
|
albertvillanova
|
[] |
Document that `Dataset.to_json` defaults to JSON Lines, by adding explanation in the corresponding docstring.
Fix #6894.
| true
|
2,292,840,226
|
https://api.github.com/repos/huggingface/datasets/issues/6894
|
https://github.com/huggingface/datasets/issues/6894
| 6,894
|
Better document defaults of to_json
|
closed
| 0
| 2024-05-13T13:30:54
| 2024-05-16T14:31:27
| 2024-05-16T14:31:27
|
albertvillanova
|
[
"documentation"
] |
Better document defaults of `to_json`: the default format is [JSON-Lines](https://jsonlines.org/).
Related to:
- #6891
| false
|
2,292,677,439
|
https://api.github.com/repos/huggingface/datasets/issues/6893
|
https://github.com/huggingface/datasets/pull/6893
| 6,893
|
Close gzipped files properly
|
closed
| 3
| 2024-05-13T12:24:39
| 2024-05-13T13:53:17
| 2024-05-13T13:01:54
|
lhoestq
|
[] |
close https://github.com/huggingface/datasets/issues/6877
| true
|
2,291,201,347
|
https://api.github.com/repos/huggingface/datasets/issues/6892
|
https://github.com/huggingface/datasets/pull/6892
| 6,892
|
Add support for categorical/dictionary types
|
closed
| 3
| 2024-05-12T07:15:08
| 2024-06-07T15:01:39
| 2024-06-07T12:20:42
|
EthanSteinberg
|
[] |
Arrow has a very useful dictionary/categorical type (https://arrow.apache.org/docs/python/generated/pyarrow.dictionary.html). This data type has significant speed, memory and disk benefits over pa.string() when there are only a few unique text strings in a column.
Unfortunately, huggingface datasets currently does not support this type. So huggingface datasets cannot natively read many parquet files that use this datatype .This PR adds support for Huggingface Datasets to read categorical/dictionary data.
Note: This PR functions by simply converting those dictionary/categorical types to strings. This means that huggingface datasets cannot take advantage of the compute benefits of categoricals, but it significantly simplifies logic. At this time, I do not think it makes sense to optimize categorical support within huggingface datasets and that we should only try to optimize later, if necessary.
Closes #5706
| true
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.